@ever-co/legal 0.4.1 → 0.5.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -0,0 +1,453 @@
1
+ <h1 id="data-protection-impact-assessment-dpia-support-pack" tabindex="-1">Data Protection Impact Assessment (DPIA) Support Pack</h1>
2
+ <p><strong>Ever Gauzy</strong> — <a href="https://gauzy.co">gauzy.co</a></p>
3
+ <p>Version 1.0.0 · In force from 2026-10-24</p>
4
+ <p>This support pack describes the processing that Ever Gauzy Cloud carries out when a customer uses it
5
+ to record working time and to monitor work. It is published by <strong>Ever Technologies LTD</strong> for the customers who
6
+ use the Service as the controller of their own people's personal data, and it is written to be copied into
7
+ their data protection impact assessment, put in front of their data protection officer or works council, and
8
+ checked by an auditor.</p>
9
+ <h2 id="whose-assessment-this-is" tabindex="-1">Whose assessment this is</h2>
10
+ <p><strong>The impact assessment is yours, not ours.</strong> Article 35 GDPR puts it on the controller, and for the data the
11
+ Service captures about your personnel the controller is you.</p>
12
+ <p>Article 28(3)(f) puts a different obligation on us as your processor: to assist you with Articles 35 and 36,
13
+ taking into account the nature of the processing and the information available to us. This pack is the main
14
+ form that assistance takes. What we commit to beyond it — answering your questionnaire, supporting a prior
15
+ consultation, telling you about changes — is set out in the impact-assessment section of our
16
+ <a href="https://gauzy.co/dpa">Data Processing Addendum</a>. That section also explains why an assessment is very
17
+ probably required before you switch monitoring on.</p>
18
+ <h2 id="what-this-pack-is-and-what-it-is-not" tabindex="-1">What this pack is, and what it is not</h2>
19
+ <ul>
20
+ <li><strong>It describes the Service as it stands on the date above</strong> — the software, its settings and their
21
+ defaults, and how we operate the hosted Service. Where we do not know an answer, it says &quot;to be confirmed&quot;
22
+ instead of guessing. Where a control is <strong>not</strong> available in Ever Gauzy today, it says so: in an
23
+ assessment, an accurate negative is worth more than a reassuring sentence.</li>
24
+ <li><strong>It is not legal advice.</strong> It does not tell you whether your processing is lawful. That depends on your
25
+ purposes, your workforce, the employment law of your country and your own balancing of interests — none of
26
+ which we can see.</li>
27
+ <li><strong>It does not amend the Data Processing Addendum.</strong> The Addendum and its Schedules remain the contract
28
+ between us. If you find this pack and the Addendum describing the same thing differently, tell us at
29
+ privacy@ever.co and we will correct whichever of them is wrong.</li>
30
+ <li><strong>It covers Ever Gauzy Cloud, the hosted Service we operate.</strong> If you run Ever Gauzy on your
31
+ own infrastructure we are not your processor for that deployment. The descriptions of the software below
32
+ still apply to it; everything about hosting, storage, sub-processors, backups and our own operations does
33
+ not, and those are yours to describe.</li>
34
+ </ul>
35
+ <h2 id="how-to-use-this-pack" tabindex="-1">How to use this pack</h2>
36
+ <p>Impact-assessment templates, including those published by supervisory authorities, ask broadly the same
37
+ questions: what is processed and about whom, how often and in what volume, for how long, where and by whom,
38
+ how it is secured, which risks it creates, and what reduces them. The sections that follow answer those
39
+ questions for Ever Gauzy:</p>
40
+ <ul>
41
+ <li><strong>where capture happens</strong> — which software captures, and which does not;</li>
42
+ <li><strong>what each feature records</strong> — feature by feature, with its default and what it does not record;</li>
43
+ <li><strong>how often, and how much</strong> — the capture intervals, and what they mean in records per person per day;</li>
44
+ <li><strong>how long data is kept, and how it is deleted</strong> — including what the product does not do automatically;</li>
45
+ <li><strong>where the data is stored, and who else receives it</strong> — storage, sub-processors and transfers;</li>
46
+ <li><strong>how it is secured</strong> — the measures that matter for monitoring data; and</li>
47
+ <li><strong>the controls, and the risk each one addresses</strong> — every privacy control the product offers, the risk it
48
+ reduces, and the risk it leaves behind.</li>
49
+ </ul>
50
+ <p><strong>Record the version and date of this pack in your assessment.</strong> The Service changes, and an assessment
51
+ should say which description it relied on. When a change could affect an assessment you have already done — a
52
+ new data category, a new capture feature, a change to a default, a new sub-processor, a change of storage
53
+ region or of a retention ceiling — we tell you, as the Data Processing Addendum provides, and we publish a new
54
+ version of this pack.</p>
55
+ <h2 id="ever-gauzy-what-is-captured-how-often-for-how-long-and-what-each-control-does" tabindex="-1">Ever Gauzy: what is captured, how often, for how long, and what each control does</h2>
56
+ <p>Everything in this part is specific to Ever Gauzy. It describes the software as it behaves on the date
57
+ of this pack, including where that behaviour is less protective than you might assume. <strong>Where a default is
58
+ &quot;on&quot;, it is on for a new organisation until an administrator changes it.</strong></p>
59
+ <h3 id="where-capture-happens" tabindex="-1">Where capture happens</h3>
60
+ <p>Capture is done by software running on a person's device, not by the platform.</p>
61
+ <ul>
62
+ <li><strong>The web application</strong> at app.gauzy.co has no screen, camera or audio capture for monitoring. Entering
63
+ time, managing tasks or reading reports in a browser produces no screenshot and no activity record.</li>
64
+ <li><strong>The desktop timer</strong> captures while the person's timer is running and stops when they stop it.</li>
65
+ <li><strong>The background agent</strong> has no timer window. It starts tracking by itself when it starts and when the
66
+ person signs in, registers itself to start when the person logs in to their computer, and shows its status
67
+ (&quot;monitoring active&quot;) in the system tray or menu bar, from where the person can stop and start tracking.
68
+ <strong>In the current version it also starts tracking again when the device wakes from sleep, even if the person
69
+ had stopped tracking before the device slept.</strong> Whether the person can quit or sign out of the agent is
70
+ decided by two settings described in the controls below.</li>
71
+ <li><strong>Connected services can bring captured data with them.</strong> Where an organisation connects Hubstaff or Upwork,
72
+ screenshots and activity levels those platforms took are imported into the workspace. Where a person
73
+ connects ActivityWatch or WakaTime, their events are added to the activity records.</li>
74
+ </ul>
75
+ <h3 id="what-each-feature-records" tabindex="-1">What each feature records</h3>
76
+ <table>
77
+ <thead>
78
+ <tr>
79
+ <th>Feature</th>
80
+ <th>What is recorded</th>
81
+ <th>Default</th>
82
+ <th>What is not recorded</th>
83
+ </tr>
84
+ </thead>
85
+ <tbody>
86
+ <tr>
87
+ <td>Time entries</td>
88
+ <td>Start and stop times, duration, project, task and client, manual entries, idle time removed or kept, and approvals</td>
89
+ <td>Core to the product</td>
90
+ <td>—</td>
91
+ </tr>
92
+ <tr>
93
+ <td>Screenshots</td>
94
+ <td>A full-resolution image of the screen and a thumbnail, linked to the time slot it belongs to. Either every connected display or the active display only. The background agent also takes one when tracking starts and when it stops or the device goes to sleep</td>
95
+ <td><strong>On.</strong> Screenshot capture is on for a new organisation and for each person, and every connected display is captured</td>
96
+ <td>The image is not blurred, masked or redacted in any way: Ever Gauzy has no blurring feature today. Nothing is captured while the desktop timer is stopped or while the agent is not tracking</td>
97
+ </tr>
98
+ <tr>
99
+ <td>Activity level, desktop timer</td>
100
+ <td>Whether the computer was idle, from the operating system's idle time. Keyboard and mouse percentages only where ActivityWatch is connected, taken from its away-from-keyboard data</td>
101
+ <td>On while the timer runs</td>
102
+ <td>The desktop timer does not listen to individual key presses and records no key identity</td>
103
+ </tr>
104
+ <tr>
105
+ <td>Activity level, background agent</td>
106
+ <td>Per one-minute interval: the number of key presses, the number of left and right mouse clicks, the number of mouse movements, and the time away from the keyboard</td>
107
+ <td><strong>Off.</strong> The keyboard and mouse tracking setting is off for a new organisation and for each person</td>
108
+ <td>With the setting off, the agent records no keyboard or mouse data</td>
109
+ </tr>
110
+ <tr>
111
+ <td>Keystroke sequence and mouse movement, background agent</td>
112
+ <td><strong>When keyboard and mouse tracking is on, the agent also records the code of every key pressed, in the order pressed, and the start and end screen coordinates of each mouse movement</strong>, and sends them to the Service, which stores them with the activity record. The API request log may keep a further copy (see retention)</td>
113
+ <td>Off, as above</td>
114
+ <td>Key codes identify physical keys, not the characters a keyboard layout produces. A sequence of them can nevertheless reveal what was typed, including passwords, so treat this as a record of keystrokes</td>
115
+ </tr>
116
+ <tr>
117
+ <td>Application usage</td>
118
+ <td>The name of the application in use and the title of its window, with the time spent. The agent samples once a second; the desktop timer every ten seconds, or uses ActivityWatch's window data where it is connected</td>
119
+ <td>Recorded whenever tracking runs. <strong>No organisation or per-person setting switches it off today</strong></td>
120
+ <td>—</td>
121
+ </tr>
122
+ <tr>
123
+ <td>Website usage</td>
124
+ <td>The address and title of the page in the active browser window: natively on macOS, and on any operating system where ActivityWatch and its browser extension are in use</td>
125
+ <td>Recorded whenever tracking runs, as above</td>
126
+ <td>—</td>
127
+ </tr>
128
+ <tr>
129
+ <td>Idle and inactivity handling, desktop timer</td>
130
+ <td>An inactivity limit after which the person is asked whether they were working, a window in which to answer, and an option to remove idle time from the timesheet</td>
131
+ <td>Inactivity tracking on; limit 10 minutes; answer window 1 minute; idle time kept</td>
132
+ <td>—</td>
133
+ </tr>
134
+ <tr>
135
+ <td>Webcam stills</td>
136
+ <td>Images from the device camera, with a thumbnail, linked to the time slot</td>
137
+ <td><strong>Not captured</strong> unless a separate camera plugin is installed on the device</td>
138
+ <td>The desktop applications themselves do not use the camera or the microphone</td>
139
+ </tr>
140
+ <tr>
141
+ <td>Audio recordings</td>
142
+ <td>Audio files recorded from the device microphone, up to 100 MB each, with their duration, channel count and sample rate. These are recordings, not sound levels</td>
143
+ <td><strong>Not captured</strong> unless a separate audio plugin is installed on the device</td>
144
+ <td>As above</td>
145
+ </tr>
146
+ <tr>
147
+ <td>Screen recordings</td>
148
+ <td>Video files of the screen, with their duration, resolution, codec and frame rate</td>
149
+ <td><strong>Not captured</strong> unless a separate video plugin is installed on the device</td>
150
+ <td>As above</td>
151
+ </tr>
152
+ <tr>
153
+ <td>ActivityWatch ingest</td>
154
+ <td>Window events (application, title, address), away-from-keyboard status, and browser events including the page address and title, whether the tab was playing sound, the number of open tabs, and <strong>whether the window was in private browsing mode</strong></td>
155
+ <td>Only where ActivityWatch is running on the device. The desktop timer's ActivityWatch option is on by default</td>
156
+ <td>Private-browsing activity is not filtered out</td>
157
+ </tr>
158
+ <tr>
159
+ <td>WakaTime ingest</td>
160
+ <td>Coding activity: editor, language, project, branch, file path and line, and the machine name</td>
161
+ <td>Only where the person connects WakaTime</td>
162
+ <td>—</td>
163
+ </tr>
164
+ <tr>
165
+ <td>AI screenshot analysis</td>
166
+ <td>For each screenshot, a generated description, the applications identified in it, and a verdict on whether the content appears to be work-related</td>
167
+ <td><strong>Off</strong> unless the organisation enables the AI integration for this purpose</td>
168
+ <td>—</td>
169
+ </tr>
170
+ <tr>
171
+ <td>Location</td>
172
+ <td>The Service does not capture a device's position during time tracking. Addresses and coordinates exist only where people enter them, on organisations, contacts and candidates</td>
173
+ <td>—</td>
174
+ <td>No GPS or device location. The API accepts a free-form location field on time slots that none of our applications fill in; an integration you build could</td>
175
+ </tr>
176
+ </tbody>
177
+ </table>
178
+ <p>For the camera, audio and video plugins, the capture behaviour on the device — interval, on-screen indicator,
179
+ which camera or microphone is used — is set by the plugin, and its description in this pack is <strong>to be
180
+ confirmed</strong>. Until it is, treat any of the three as unassessed and keep them off.</p>
181
+ <h3 id="how-often-and-how-much" tabindex="-1">How often, and how much</h3>
182
+ <p><strong>Screenshot intervals.</strong> The available intervals are 1, 3, 5 and 10 minutes. There is no longer interval.</p>
183
+ <ul>
184
+ <li><strong>Desktop timer.</strong> The interval is a setting on the device, 10 minutes by default, which the person can
185
+ change while the timer is stopped. The organisation's screenshot interval, randomised capture and
186
+ capture-on-sleep settings are applied to the device <strong>only when the organisation's timer policy is
187
+ enforced</strong>, which it is not by default.</li>
188
+ <li><strong>Background agent.</strong> The interval comes from the agent's own settings on the device. <strong>The agent does not
189
+ read the organisation's screenshot interval.</strong></li>
190
+ </ul>
191
+ <p><strong>Randomised capture</strong> is off by default. It works differently in the two applications:</p>
192
+ <ul>
193
+ <li>in the desktop timer, each screenshot is taken at a random point that is never later than the configured
194
+ interval, so gaps range from 1 minute up to the interval;</li>
195
+ <li>in the background agent, each next gap is chosen at random from 1, 3, 5 and 10 minutes <strong>regardless of
196
+ the configured interval</strong>, an average of 4.75 minutes.</li>
197
+ </ul>
198
+ <p><strong>What that means per person.</strong> For a person tracked for 8 hours, with one display:</p>
199
+ <table>
200
+ <thead>
201
+ <tr>
202
+ <th>Configuration</th>
203
+ <th>Screenshots in 8 hours</th>
204
+ </tr>
205
+ </thead>
206
+ <tbody>
207
+ <tr>
208
+ <td>Fixed, every 10 minutes</td>
209
+ <td>48</td>
210
+ </tr>
211
+ <tr>
212
+ <td>Fixed, every 5 minutes</td>
213
+ <td>96</td>
214
+ </tr>
215
+ <tr>
216
+ <td>Fixed, every 3 minutes</td>
217
+ <td>160</td>
218
+ </tr>
219
+ <tr>
220
+ <td>Fixed, every minute</td>
221
+ <td>480</td>
222
+ </tr>
223
+ <tr>
224
+ <td>Randomised, desktop timer, 10-minute interval</td>
225
+ <td>about 80 on average, and up to 480</td>
226
+ </tr>
227
+ <tr>
228
+ <td>Randomised, background agent, any interval</td>
229
+ <td>about 100 on average, and up to 480</td>
230
+ </tr>
231
+ </tbody>
232
+ </table>
233
+ <p>Multiply by the number of displays where every display is captured, which is the default. The background agent
234
+ adds one screenshot each time tracking starts and each time it stops or the device sleeps.</p>
235
+ <p><strong>Activity records.</strong> The background agent sends an activity record every minute while it tracks — 480 in an
236
+ 8-hour day — and samples the active window every second. The desktop timer samples the active window every ten
237
+ seconds and sends its records at each capture interval.</p>
238
+ <h3 id="how-long-it-is-kept-and-how-it-is-deleted" tabindex="-1">How long it is kept, and how it is deleted</h3>
239
+ <p><strong>Ever Gauzy does not delete monitoring data automatically today.</strong> It has no retention setting for
240
+ screenshots, other captured media, activity records or application and website records, and no scheduled job
241
+ that deletes them. They are kept until someone deletes them, or until your subscription ends and the deletion
242
+ timetable in the Data Processing Addendum runs. If your assessment sets a retention period, you implement it
243
+ by deleting the data on that schedule, or by asking us to.</p>
244
+ <p><strong>Deleting inside the application is a soft deletion by default.</strong> Deleting a screenshot, a time slot or a
245
+ time entry in the web application marks the record as deleted and removes it from timesheets, galleries and
246
+ reports. The database record stays, and so do the image and its thumbnail in object storage. Permanently
247
+ deleting a screenshot through the API removes its record and deletes the stored image and thumbnail. If you need
248
+ soft-deleted captures removed permanently and cannot do it yourself, write to privacy@ever.co and we will
249
+ treat it as a deletion instruction under the Data Processing Addendum.</p>
250
+ <p><strong>A request log holds a further copy.</strong> The API keeps a log of the time-tracking requests it receives — the
251
+ address, the IP address and user agent, and the request and response bodies, with passwords and tokens
252
+ redacted — which therefore includes activity data sent by the desktop applications, keystroke sequences
253
+ included where they are recorded. The log is on by default in the software and has no automatic expiry.
254
+ Whether it is enabled on Ever Gauzy Cloud, and for how long its entries are kept, is <strong>to be
255
+ confirmed</strong>.</p>
256
+ <p><strong>At the end of your subscription</strong> you have 30 days to export your data, and deletion from our live systems
257
+ then completes within a further 30 days. <strong>Backups</strong> age out within 30 days of a deletion. Both are set out in
258
+ the Data Processing Addendum and the retention section of our <a href="https://gauzy.co/privacy">Privacy Policy</a>.</p>
259
+ <h3 id="where-monitoring-data-is-stored-on-ever-gauzy-cloud" tabindex="-1">Where monitoring data is stored on Ever Gauzy Cloud</h3>
260
+ <ul>
261
+ <li><strong>Records</strong> — time slots, activity records, screenshot metadata, settings — are held in the PostgreSQL
262
+ database we operate.</li>
263
+ <li><strong>Captured files</strong> — screenshots, thumbnails and any camera, audio or video files — are held in
264
+ S3-compatible object storage that we operate, on our own infrastructure in the European Union. They are not
265
+ public: the applications fetch them through signed links that expire after one hour.</li>
266
+ <li><strong>A workspace can be configured to use its own storage provider</strong> instead. Where your organisation does
267
+ that, captured files go to that provider under your own agreement with it.</li>
268
+ <li><strong>Where AI screenshot analysis is enabled</strong>, each screenshot is sent to the Gauzy AI service for analysis.
269
+ Where that service runs, and when content goes onward to a model provider, is described in the annex to our
270
+ <a href="https://gauzy.co/privacy">Privacy Policy</a>.</li>
271
+ <li><strong>Customer separation</strong> is logical. All workspaces share the application and the database; every record
272
+ carries the tenant and organisation it belongs to, and the application uses them to scope queries to the
273
+ workspace of the person asking.</li>
274
+ </ul>
275
+ <h3 id="the-controls-and-the-risk-each-one-addresses" tabindex="-1">The controls, and the risk each one addresses</h3>
276
+ <table>
277
+ <thead>
278
+ <tr>
279
+ <th>Control</th>
280
+ <th>Default</th>
281
+ <th>Risk it reduces</th>
282
+ <th>What it leaves, and what you still need to do</th>
283
+ </tr>
284
+ </thead>
285
+ <tbody>
286
+ <tr>
287
+ <td>Screenshots off, for the organisation or for one person</td>
288
+ <td>Screenshots on</td>
289
+ <td>Systematic visual monitoring, and incidental capture of special-category data, private messages and third parties on screen</td>
290
+ <td>While screenshots are on, everything visible is captured at full resolution. Switch them off for people and roles you have not assessed as needing them</td>
291
+ </tr>
292
+ <tr>
293
+ <td>Active display only, instead of every display</td>
294
+ <td>Every display</td>
295
+ <td>Capture of screens the work is not on, and multiplied volume</td>
296
+ <td>The active display can still show anything</td>
297
+ </tr>
298
+ <tr>
299
+ <td>Capture interval</td>
300
+ <td>10 minutes</td>
301
+ <td>Volume of capture, under data minimisation</td>
302
+ <td>10 minutes is the longest interval available. On the desktop timer the organisation's interval applies only when the timer policy is enforced; the agent does not read it</td>
303
+ </tr>
304
+ <tr>
305
+ <td>Enforcing the organisation's timer policy</td>
306
+ <td>Not enforced</td>
307
+ <td>Inconsistent per-device settings</td>
308
+ <td>Enforcing also removes per-person exceptions. It does not reach the background agent's interval</td>
309
+ </tr>
310
+ <tr>
311
+ <td>Randomised capture</td>
312
+ <td>Off</td>
313
+ <td>Predictable capture that people work around</td>
314
+ <td>Changes volume, in a direction that depends on the application and the interval. In the desktop timer it always increases it — about 80 screenshots in 8 hours at a 10-minute interval instead of 48. In the agent it gives about 100 in 8 hours whatever the interval: more than a fixed 5 or 10 minutes (96 or 48), fewer than a fixed 1 or 3 minutes (480 or 160)</td>
315
+ </tr>
316
+ <tr>
317
+ <td>Keyboard and mouse tracking, background agent</td>
318
+ <td>Off</td>
319
+ <td>With it off, no keyboard or mouse data and no keystroke sequence is recorded</td>
320
+ <td><strong>With it on, the sequence of keys pressed and mouse coordinates are recorded.</strong> We consider keeping it off load-bearing for a defensible configuration</td>
321
+ </tr>
322
+ <tr>
323
+ <td>Inactivity limit and removal of idle time, desktop timer</td>
324
+ <td>Limit 10 minutes, answer window 1 minute, idle time kept</td>
325
+ <td>Time recorded while nobody is working</td>
326
+ <td>Idle detection is by input and operating-system idle time, not by what the person is doing</td>
327
+ </tr>
328
+ <tr>
329
+ <td>Visibility of tracked data to the person tracked</td>
330
+ <td>On</td>
331
+ <td>People not knowing what is recorded about them; supports their right of access</td>
332
+ <td>The API enforces the setting. It does not replace the notice you owe your people</td>
333
+ </tr>
334
+ <tr>
335
+ <td>Agent exit and sign-out</td>
336
+ <td>Allowed</td>
337
+ <td>Monitoring the person cannot stop</td>
338
+ <td>The API refuses to restrict either setting for a person or organisation whose country, locale or time zone on record is in the EEA or the UK. Elsewhere a restriction needs an explicit acknowledgement, which is recorded against the administrator's account. A person with no location or time zone on record is treated as outside the EEA and UK, and restrictions applied before this check existed are not lifted automatically</td>
339
+ </tr>
340
+ <tr>
341
+ <td>Deleting a capture</td>
342
+ <td>—</td>
343
+ <td>A capture that caught something it should not have</td>
344
+ <td>In the web application deletion is soft: the stored image remains until it is permanently deleted</td>
345
+ </tr>
346
+ <tr>
347
+ <td>Camera, audio and screen recording plugins</td>
348
+ <td>Not installed</td>
349
+ <td>The most intrusive capture types</td>
350
+ <td>Each needs a separate decision and its own assessment</td>
351
+ </tr>
352
+ <tr>
353
+ <td>AI screenshot analysis</td>
354
+ <td>Off</td>
355
+ <td>—</td>
356
+ <td>When on, an automated verdict on whether content appears work-related is attached to each screenshot. It is an input for a person to weigh, not a decision</td>
357
+ </tr>
358
+ </tbody>
359
+ </table>
360
+ <p><strong>Controls that Ever Gauzy does not offer today.</strong> No blurring or masking of screenshots; no excluded
361
+ applications or websites; no per-person pause other than stopping the timer or the agent's tracking; no
362
+ retention period with automatic deletion; and no log of who viewed a capture — the activity log records
363
+ creation, changes and deletions, not viewing. If your assessment depends on any of these, it cannot rely on
364
+ Ever Gauzy to provide it today.</p>
365
+ <h3 id="what-we-can-tell-you-about-your-own-account" tabindex="-1">What we can tell you about your own account</h3>
366
+ <p>On request at privacy@ever.co we will tell you, for your assessment: which capture features are enabled
367
+ in your workspace and their current settings, where its captured files are stored, and which sub-processors
368
+ are involved.</p>
369
+ <h2 id="hosting-sub-processors-and-transfers" tabindex="-1">Hosting, sub-processors and transfers</h2>
370
+ <p>This section summarises what our <a href="https://gauzy.co/subprocessors">sub-processor page</a> and
371
+ <a href="https://gauzy.co/security">Security page</a> say in full. Where this summary and those pages differ, those pages
372
+ are the current statement and this summary is the one to correct.</p>
373
+ <h3 id="where-processing-happens" tabindex="-1">Where processing happens</h3>
374
+ <ul>
375
+ <li><strong>Primary processing is in the European Union, on infrastructure we operate.</strong> The Service runs on physical
376
+ servers we own, in facilities we control, inside the European Union. The database and the object storage
377
+ that hold Customer Personal Data are components we run, not services bought from a third party.</li>
378
+ <li><strong>Requests pass through a content delivery and security network</strong> that terminates traffic at the edge
379
+ location nearest to whoever makes the request, which can be anywhere in the world, before carrying it to our
380
+ infrastructure. That provider is in the always-engaged tier of the sub-processor list.</li>
381
+ <li><strong>Off-site backup copies are held by an external object storage provider</strong>, encrypted by us before they
382
+ leave our network. The provider holds ciphertext and no key.</li>
383
+ <li><strong>Our source code, build pipeline and container images are hosted by a third party.</strong> That is how the
384
+ Service is built and deployed, not where your data is stored.</li>
385
+ </ul>
386
+ <p>This is not multi-region hosting, and we do not describe it as such. The continuity section of the Security
387
+ page explains what a site-level failure would mean.</p>
388
+ <h3 id="sub-processors" tabindex="-1">Sub-processors</h3>
389
+ <p>The <a href="https://gauzy.co/subprocessors">sub-processor list</a> for Ever Gauzy has three tiers, and the tier
390
+ matters as much as the name:</p>
391
+ <ul>
392
+ <li><strong>Tier 1 — always engaged.</strong> In the path for every customer, with no setting that removes them. This is the
393
+ Article 28 list proper.</li>
394
+ <li><strong>Tier 2 — engaged only when something brings them in</strong>: a feature an administrator enables, an
395
+ integration someone connects, a provider configured for the workspace, or a cookie category a visitor
396
+ accepts. Several of them are providers that you contract with directly, under your own account.</li>
397
+ <li><strong>Tier 3 — self-hosted deployments only.</strong> Providers the software can be configured to use in a deployment
398
+ you run. We are not a processor for anything in that tier.</li>
399
+ </ul>
400
+ <p>For each provider the list gives the purpose, the location, the categories of personal data it receives and
401
+ the transfer mechanism relied on. For your assessment, read the categories of data column against the
402
+ features you actually enable: a tier 2 provider is not processing your data unless the condition next to it is
403
+ true.</p>
404
+ <h3 id="transfers-outside-the-european-economic-area" tabindex="-1">Transfers outside the European Economic Area</h3>
405
+ <p>Where a provider processes personal data outside the European Economic Area, the transfer mechanism is stated
406
+ against it on the sub-processor list, and Schedule 3 of the Data Processing Addendum describes the mechanism
407
+ and the transfer impact assessment behind it. <strong>Where we cannot yet evidence a mechanism for a provider, the
408
+ list says &quot;To be confirmed&quot; rather than naming one.</strong> Read those rows as an open item, and ask us at
409
+ privacy@ever.co for the current position on any provider that matters to your assessment.</p>
410
+ <p>If you need a copy of the safeguards for a named provider — the Standard Contractual Clauses and the modules
411
+ that apply — write to privacy@ever.co and we will send them.</p>
412
+ <h2 id="security-measures" tabindex="-1">Security measures</h2>
413
+ <p>The technical and organisational measures we operate are described on our
414
+ <a href="https://gauzy.co/security">Security page</a> and, contractually, in Schedule 2 of the Data Processing Addendum.
415
+ The points below are the ones an assessment of monitoring data usually turns on.</p>
416
+ <ul>
417
+ <li><strong>Encryption.</strong> Traffic between users and the Service travels over TLS, and plain HTTP is redirected rather
418
+ than served. Storage holding customer content, databases and backups is encrypted at rest, and off-site
419
+ backup copies are encrypted before they leave our network, with keys we hold. Encryption at rest protects
420
+ against the loss or theft of physical media; it does not protect data from a running system that has
421
+ already decrypted it.</li>
422
+ <li><strong>Access by our personnel.</strong> Production access uses named individual accounts, is scoped to the narrowest
423
+ role that lets the work happen, and is revoked when someone no longer needs it. We access the content of a
424
+ workspace only where that is necessary to operate, secure or support the Service, and we do not browse or
425
+ review captured data for purposes of our own. The access-control section of the Security page describes
426
+ this, including its limits for a team of our size.</li>
427
+ <li><strong>Separation between customers is logical, not physical</strong>, unless you have separately contracted for a
428
+ dedicated deployment. The product-specific part of this pack states how that separation works in
429
+ Ever Gauzy.</li>
430
+ <li><strong>Backups and continuity.</strong> Backups are held in three tiers, the last of them off-site and encrypted, and
431
+ each backup ages out on its own rolling cycle within the period stated in the retention section of our
432
+ <a href="https://gauzy.co/privacy">Privacy Policy</a>.</li>
433
+ </ul>
434
+ <p><strong>We do not hold a SOC 2 report of any type, and we are not certified to ISO/IEC 27001.</strong> If any document has
435
+ told you otherwise, it is wrong. What we offer instead — the Security page, Schedule 2, answers to a security
436
+ questionnaire and evidence on request — is described on the Security page.</p>
437
+ <h2 id="questions-and-a-version-you-can-cite" tabindex="-1">Questions, and a version you can cite</h2>
438
+ <ul>
439
+ <li><strong>A question about anything in this pack, an impact-assessment questionnaire, or a request for the current
440
+ settings of your own account</strong> — privacy@ever.co</li>
441
+ <li><strong>A request for the Data Processing Addendum, a security questionnaire, or evidence of a restore test</strong> —
442
+ legal@ever.co</li>
443
+ <li><strong>Something you think has been compromised — tell us immediately</strong> — abuse@ever.co</li>
444
+ </ul>
445
+ <p>By post: <strong>Ever Technologies LTD</strong>, Mladost 2, bl. 211, ent. A, Sofia 1799, Bulgaria — registered in Bulgaria under company number
446
+ 204599535. We correspond in English.</p>
447
+ <p><strong>We have not designated a Data Protection Officer under Article 37 GDPR.</strong> dpo@ever.co is a contact
448
+ channel, monitored by the same people as privacy@ever.co, and we will not describe it as a designation. If
449
+ a questionnaire asks for our data protection officer, that sentence is the honest answer.</p>
450
+ <p>This document is version 1.0.0 of the DPIA Support Pack for
451
+ <a href="https://gauzy.co">gauzy.co</a>, in force from 2026-10-24. It describes the Service as it
452
+ stood on that date. Cite the version and the date in your assessment, and check for a newer version before
453
+ you rely on it again.</p>
@@ -0,0 +1,97 @@
1
+ {
2
+ "document": "dpia",
3
+ "product": "gauzy",
4
+ "productName": "Ever Gauzy",
5
+ "domain": "gauzy.co",
6
+ "entity": "Ever Technologies LTD",
7
+ "entityId": "ever-technologies-bg",
8
+ "locale": "en",
9
+ "version": "1.0.0",
10
+ "effectiveDate": "2026-10-24",
11
+ "sha256": "20b37e6a3140a5c11cf3c82230930f0ece5b42f422b3771fda3a787558a01abf",
12
+ "publishable": true,
13
+ "sections": [
14
+ "core/dpia/010-purpose-and-status",
15
+ "core/dpia/020-how-to-use-this-pack",
16
+ "@annex/dpia",
17
+ "core/dpia/100-hosting-subprocessors-and-transfers",
18
+ "core/dpia/110-security-measures",
19
+ "core/dpia/900-contact"
20
+ ],
21
+ "headings": [
22
+ {
23
+ "level": 1,
24
+ "text": "Data Protection Impact Assessment (DPIA) Support Pack"
25
+ },
26
+ {
27
+ "level": 2,
28
+ "text": "Whose assessment this is"
29
+ },
30
+ {
31
+ "level": 2,
32
+ "text": "What this pack is, and what it is not"
33
+ },
34
+ {
35
+ "level": 2,
36
+ "text": "How to use this pack"
37
+ },
38
+ {
39
+ "level": 2,
40
+ "text": "Ever Gauzy: what is captured, how often, for how long, and what each control does"
41
+ },
42
+ {
43
+ "level": 3,
44
+ "text": "Where capture happens"
45
+ },
46
+ {
47
+ "level": 3,
48
+ "text": "What each feature records"
49
+ },
50
+ {
51
+ "level": 3,
52
+ "text": "How often, and how much"
53
+ },
54
+ {
55
+ "level": 3,
56
+ "text": "How long it is kept, and how it is deleted"
57
+ },
58
+ {
59
+ "level": 3,
60
+ "text": "Where monitoring data is stored on Ever Gauzy Cloud"
61
+ },
62
+ {
63
+ "level": 3,
64
+ "text": "The controls, and the risk each one addresses"
65
+ },
66
+ {
67
+ "level": 3,
68
+ "text": "What we can tell you about your own account"
69
+ },
70
+ {
71
+ "level": 2,
72
+ "text": "Hosting, sub-processors and transfers"
73
+ },
74
+ {
75
+ "level": 3,
76
+ "text": "Where processing happens"
77
+ },
78
+ {
79
+ "level": 3,
80
+ "text": "Sub-processors"
81
+ },
82
+ {
83
+ "level": 3,
84
+ "text": "Transfers outside the European Economic Area"
85
+ },
86
+ {
87
+ "level": 2,
88
+ "text": "Security measures"
89
+ },
90
+ {
91
+ "level": 2,
92
+ "text": "Questions, and a version you can cite"
93
+ }
94
+ ],
95
+ "path": "content/dpia/gauzy/en.html",
96
+ "markdown": "# Data Protection Impact Assessment (DPIA) Support Pack\n\n**Ever Gauzy** — [gauzy.co](https://gauzy.co)\n\nVersion 1.0.0 · In force from 2026-10-24\n\nThis support pack describes the processing that Ever Gauzy Cloud carries out when a customer uses it\nto record working time and to monitor work. It is published by **Ever Technologies LTD** for the customers who\nuse the Service as the controller of their own people's personal data, and it is written to be copied into\ntheir data protection impact assessment, put in front of their data protection officer or works council, and\nchecked by an auditor.\n\n## Whose assessment this is\n\n**The impact assessment is yours, not ours.** Article 35 GDPR puts it on the controller, and for the data the\nService captures about your personnel the controller is you.\n\nArticle 28(3)(f) puts a different obligation on us as your processor: to assist you with Articles 35 and 36,\ntaking into account the nature of the processing and the information available to us. This pack is the main\nform that assistance takes. What we commit to beyond it — answering your questionnaire, supporting a prior\nconsultation, telling you about changes — is set out in the impact-assessment section of our\n[Data Processing Addendum](https://gauzy.co/dpa). That section also explains why an assessment is very\nprobably required before you switch monitoring on.\n\n## What this pack is, and what it is not\n\n- **It describes the Service as it stands on the date above** — the software, its settings and their\n defaults, and how we operate the hosted Service. Where we do not know an answer, it says \"to be confirmed\"\n instead of guessing. Where a control is **not** available in Ever Gauzy today, it says so: in an\n assessment, an accurate negative is worth more than a reassuring sentence.\n- **It is not legal advice.** It does not tell you whether your processing is lawful. That depends on your\n purposes, your workforce, the employment law of your country and your own balancing of interests — none of\n which we can see.\n- **It does not amend the Data Processing Addendum.** The Addendum and its Schedules remain the contract\n between us. If you find this pack and the Addendum describing the same thing differently, tell us at\n privacy@ever.co and we will correct whichever of them is wrong.\n- **It covers Ever Gauzy Cloud, the hosted Service we operate.** If you run Ever Gauzy on your\n own infrastructure we are not your processor for that deployment. The descriptions of the software below\n still apply to it; everything about hosting, storage, sub-processors, backups and our own operations does\n not, and those are yours to describe.\n\n## How to use this pack\n\nImpact-assessment templates, including those published by supervisory authorities, ask broadly the same\nquestions: what is processed and about whom, how often and in what volume, for how long, where and by whom,\nhow it is secured, which risks it creates, and what reduces them. The sections that follow answer those\nquestions for Ever Gauzy:\n\n- **where capture happens** — which software captures, and which does not;\n- **what each feature records** — feature by feature, with its default and what it does not record;\n- **how often, and how much** — the capture intervals, and what they mean in records per person per day;\n- **how long data is kept, and how it is deleted** — including what the product does not do automatically;\n- **where the data is stored, and who else receives it** — storage, sub-processors and transfers;\n- **how it is secured** — the measures that matter for monitoring data; and\n- **the controls, and the risk each one addresses** — every privacy control the product offers, the risk it\n reduces, and the risk it leaves behind.\n\n**Record the version and date of this pack in your assessment.** The Service changes, and an assessment\nshould say which description it relied on. When a change could affect an assessment you have already done — a\nnew data category, a new capture feature, a change to a default, a new sub-processor, a change of storage\nregion or of a retention ceiling — we tell you, as the Data Processing Addendum provides, and we publish a new\nversion of this pack.\n\n## Ever Gauzy: what is captured, how often, for how long, and what each control does\n\nEverything in this part is specific to Ever Gauzy. It describes the software as it behaves on the date\nof this pack, including where that behaviour is less protective than you might assume. **Where a default is\n\"on\", it is on for a new organisation until an administrator changes it.**\n\n### Where capture happens\n\nCapture is done by software running on a person's device, not by the platform.\n\n- **The web application** at app.gauzy.co has no screen, camera or audio capture for monitoring. Entering\n time, managing tasks or reading reports in a browser produces no screenshot and no activity record.\n- **The desktop timer** captures while the person's timer is running and stops when they stop it.\n- **The background agent** has no timer window. It starts tracking by itself when it starts and when the\n person signs in, registers itself to start when the person logs in to their computer, and shows its status\n (\"monitoring active\") in the system tray or menu bar, from where the person can stop and start tracking.\n **In the current version it also starts tracking again when the device wakes from sleep, even if the person\n had stopped tracking before the device slept.** Whether the person can quit or sign out of the agent is\n decided by two settings described in the controls below.\n- **Connected services can bring captured data with them.** Where an organisation connects Hubstaff or Upwork,\n screenshots and activity levels those platforms took are imported into the workspace. Where a person\n connects ActivityWatch or WakaTime, their events are added to the activity records.\n\n### What each feature records\n\n| Feature | What is recorded | Default | What is not recorded |\n|---|---|---|---|\n| Time entries | Start and stop times, duration, project, task and client, manual entries, idle time removed or kept, and approvals | Core to the product | — |\n| Screenshots | A full-resolution image of the screen and a thumbnail, linked to the time slot it belongs to. Either every connected display or the active display only. The background agent also takes one when tracking starts and when it stops or the device goes to sleep | **On.** Screenshot capture is on for a new organisation and for each person, and every connected display is captured | The image is not blurred, masked or redacted in any way: Ever Gauzy has no blurring feature today. Nothing is captured while the desktop timer is stopped or while the agent is not tracking |\n| Activity level, desktop timer | Whether the computer was idle, from the operating system's idle time. Keyboard and mouse percentages only where ActivityWatch is connected, taken from its away-from-keyboard data | On while the timer runs | The desktop timer does not listen to individual key presses and records no key identity |\n| Activity level, background agent | Per one-minute interval: the number of key presses, the number of left and right mouse clicks, the number of mouse movements, and the time away from the keyboard | **Off.** The keyboard and mouse tracking setting is off for a new organisation and for each person | With the setting off, the agent records no keyboard or mouse data |\n| Keystroke sequence and mouse movement, background agent | **When keyboard and mouse tracking is on, the agent also records the code of every key pressed, in the order pressed, and the start and end screen coordinates of each mouse movement**, and sends them to the Service, which stores them with the activity record. The API request log may keep a further copy (see retention) | Off, as above | Key codes identify physical keys, not the characters a keyboard layout produces. A sequence of them can nevertheless reveal what was typed, including passwords, so treat this as a record of keystrokes |\n| Application usage | The name of the application in use and the title of its window, with the time spent. The agent samples once a second; the desktop timer every ten seconds, or uses ActivityWatch's window data where it is connected | Recorded whenever tracking runs. **No organisation or per-person setting switches it off today** | — |\n| Website usage | The address and title of the page in the active browser window: natively on macOS, and on any operating system where ActivityWatch and its browser extension are in use | Recorded whenever tracking runs, as above | — |\n| Idle and inactivity handling, desktop timer | An inactivity limit after which the person is asked whether they were working, a window in which to answer, and an option to remove idle time from the timesheet | Inactivity tracking on; limit 10 minutes; answer window 1 minute; idle time kept | — |\n| Webcam stills | Images from the device camera, with a thumbnail, linked to the time slot | **Not captured** unless a separate camera plugin is installed on the device | The desktop applications themselves do not use the camera or the microphone |\n| Audio recordings | Audio files recorded from the device microphone, up to 100 MB each, with their duration, channel count and sample rate. These are recordings, not sound levels | **Not captured** unless a separate audio plugin is installed on the device | As above |\n| Screen recordings | Video files of the screen, with their duration, resolution, codec and frame rate | **Not captured** unless a separate video plugin is installed on the device | As above |\n| ActivityWatch ingest | Window events (application, title, address), away-from-keyboard status, and browser events including the page address and title, whether the tab was playing sound, the number of open tabs, and **whether the window was in private browsing mode** | Only where ActivityWatch is running on the device. The desktop timer's ActivityWatch option is on by default | Private-browsing activity is not filtered out |\n| WakaTime ingest | Coding activity: editor, language, project, branch, file path and line, and the machine name | Only where the person connects WakaTime | — |\n| AI screenshot analysis | For each screenshot, a generated description, the applications identified in it, and a verdict on whether the content appears to be work-related | **Off** unless the organisation enables the AI integration for this purpose | — |\n| Location | The Service does not capture a device's position during time tracking. Addresses and coordinates exist only where people enter them, on organisations, contacts and candidates | — | No GPS or device location. The API accepts a free-form location field on time slots that none of our applications fill in; an integration you build could |\n\nFor the camera, audio and video plugins, the capture behaviour on the device — interval, on-screen indicator,\nwhich camera or microphone is used — is set by the plugin, and its description in this pack is **to be\nconfirmed**. Until it is, treat any of the three as unassessed and keep them off.\n\n### How often, and how much\n\n**Screenshot intervals.** The available intervals are 1, 3, 5 and 10 minutes. There is no longer interval.\n\n- **Desktop timer.** The interval is a setting on the device, 10 minutes by default, which the person can\n change while the timer is stopped. The organisation's screenshot interval, randomised capture and\n capture-on-sleep settings are applied to the device **only when the organisation's timer policy is\n enforced**, which it is not by default.\n- **Background agent.** The interval comes from the agent's own settings on the device. **The agent does not\n read the organisation's screenshot interval.**\n\n**Randomised capture** is off by default. It works differently in the two applications:\n\n- in the desktop timer, each screenshot is taken at a random point that is never later than the configured\n interval, so gaps range from 1 minute up to the interval;\n- in the background agent, each next gap is chosen at random from 1, 3, 5 and 10 minutes **regardless of\n the configured interval**, an average of 4.75 minutes.\n\n**What that means per person.** For a person tracked for 8 hours, with one display:\n\n| Configuration | Screenshots in 8 hours |\n|---|---|\n| Fixed, every 10 minutes | 48 |\n| Fixed, every 5 minutes | 96 |\n| Fixed, every 3 minutes | 160 |\n| Fixed, every minute | 480 |\n| Randomised, desktop timer, 10-minute interval | about 80 on average, and up to 480 |\n| Randomised, background agent, any interval | about 100 on average, and up to 480 |\n\nMultiply by the number of displays where every display is captured, which is the default. The background agent\nadds one screenshot each time tracking starts and each time it stops or the device sleeps.\n\n**Activity records.** The background agent sends an activity record every minute while it tracks — 480 in an\n8-hour day — and samples the active window every second. The desktop timer samples the active window every ten\nseconds and sends its records at each capture interval.\n\n### How long it is kept, and how it is deleted\n\n**Ever Gauzy does not delete monitoring data automatically today.** It has no retention setting for\nscreenshots, other captured media, activity records or application and website records, and no scheduled job\nthat deletes them. They are kept until someone deletes them, or until your subscription ends and the deletion\ntimetable in the Data Processing Addendum runs. If your assessment sets a retention period, you implement it\nby deleting the data on that schedule, or by asking us to.\n\n**Deleting inside the application is a soft deletion by default.** Deleting a screenshot, a time slot or a\ntime entry in the web application marks the record as deleted and removes it from timesheets, galleries and\nreports. The database record stays, and so do the image and its thumbnail in object storage. Permanently\ndeleting a screenshot through the API removes its record and deletes the stored image and thumbnail. If you need\nsoft-deleted captures removed permanently and cannot do it yourself, write to privacy@ever.co and we will\ntreat it as a deletion instruction under the Data Processing Addendum.\n\n**A request log holds a further copy.** The API keeps a log of the time-tracking requests it receives — the\naddress, the IP address and user agent, and the request and response bodies, with passwords and tokens\nredacted — which therefore includes activity data sent by the desktop applications, keystroke sequences\nincluded where they are recorded. The log is on by default in the software and has no automatic expiry.\nWhether it is enabled on Ever Gauzy Cloud, and for how long its entries are kept, is **to be\nconfirmed**.\n\n**At the end of your subscription** you have 30 days to export your data, and deletion from our live systems\nthen completes within a further 30 days. **Backups** age out within 30 days of a deletion. Both are set out in\nthe Data Processing Addendum and the retention section of our [Privacy Policy](https://gauzy.co/privacy).\n\n### Where monitoring data is stored on Ever Gauzy Cloud\n\n- **Records** — time slots, activity records, screenshot metadata, settings — are held in the PostgreSQL\n database we operate.\n- **Captured files** — screenshots, thumbnails and any camera, audio or video files — are held in\n S3-compatible object storage that we operate, on our own infrastructure in the European Union. They are not\n public: the applications fetch them through signed links that expire after one hour.\n- **A workspace can be configured to use its own storage provider** instead. Where your organisation does\n that, captured files go to that provider under your own agreement with it.\n- **Where AI screenshot analysis is enabled**, each screenshot is sent to the Gauzy AI service for analysis.\n Where that service runs, and when content goes onward to a model provider, is described in the annex to our\n [Privacy Policy](https://gauzy.co/privacy).\n- **Customer separation** is logical. All workspaces share the application and the database; every record\n carries the tenant and organisation it belongs to, and the application uses them to scope queries to the\n workspace of the person asking.\n\n### The controls, and the risk each one addresses\n\n| Control | Default | Risk it reduces | What it leaves, and what you still need to do |\n|---|---|---|---|\n| Screenshots off, for the organisation or for one person | Screenshots on | Systematic visual monitoring, and incidental capture of special-category data, private messages and third parties on screen | While screenshots are on, everything visible is captured at full resolution. Switch them off for people and roles you have not assessed as needing them |\n| Active display only, instead of every display | Every display | Capture of screens the work is not on, and multiplied volume | The active display can still show anything |\n| Capture interval | 10 minutes | Volume of capture, under data minimisation | 10 minutes is the longest interval available. On the desktop timer the organisation's interval applies only when the timer policy is enforced; the agent does not read it |\n| Enforcing the organisation's timer policy | Not enforced | Inconsistent per-device settings | Enforcing also removes per-person exceptions. It does not reach the background agent's interval |\n| Randomised capture | Off | Predictable capture that people work around | Changes volume, in a direction that depends on the application and the interval. In the desktop timer it always increases it — about 80 screenshots in 8 hours at a 10-minute interval instead of 48. In the agent it gives about 100 in 8 hours whatever the interval: more than a fixed 5 or 10 minutes (96 or 48), fewer than a fixed 1 or 3 minutes (480 or 160) |\n| Keyboard and mouse tracking, background agent | Off | With it off, no keyboard or mouse data and no keystroke sequence is recorded | **With it on, the sequence of keys pressed and mouse coordinates are recorded.** We consider keeping it off load-bearing for a defensible configuration |\n| Inactivity limit and removal of idle time, desktop timer | Limit 10 minutes, answer window 1 minute, idle time kept | Time recorded while nobody is working | Idle detection is by input and operating-system idle time, not by what the person is doing |\n| Visibility of tracked data to the person tracked | On | People not knowing what is recorded about them; supports their right of access | The API enforces the setting. It does not replace the notice you owe your people |\n| Agent exit and sign-out | Allowed | Monitoring the person cannot stop | The API refuses to restrict either setting for a person or organisation whose country, locale or time zone on record is in the EEA or the UK. Elsewhere a restriction needs an explicit acknowledgement, which is recorded against the administrator's account. A person with no location or time zone on record is treated as outside the EEA and UK, and restrictions applied before this check existed are not lifted automatically |\n| Deleting a capture | — | A capture that caught something it should not have | In the web application deletion is soft: the stored image remains until it is permanently deleted |\n| Camera, audio and screen recording plugins | Not installed | The most intrusive capture types | Each needs a separate decision and its own assessment |\n| AI screenshot analysis | Off | — | When on, an automated verdict on whether content appears work-related is attached to each screenshot. It is an input for a person to weigh, not a decision |\n\n**Controls that Ever Gauzy does not offer today.** No blurring or masking of screenshots; no excluded\napplications or websites; no per-person pause other than stopping the timer or the agent's tracking; no\nretention period with automatic deletion; and no log of who viewed a capture — the activity log records\ncreation, changes and deletions, not viewing. If your assessment depends on any of these, it cannot rely on\nEver Gauzy to provide it today.\n\n### What we can tell you about your own account\n\nOn request at privacy@ever.co we will tell you, for your assessment: which capture features are enabled\nin your workspace and their current settings, where its captured files are stored, and which sub-processors\nare involved.\n\n## Hosting, sub-processors and transfers\n\nThis section summarises what our [sub-processor page](https://gauzy.co/subprocessors) and\n[Security page](https://gauzy.co/security) say in full. Where this summary and those pages differ, those pages\nare the current statement and this summary is the one to correct.\n\n### Where processing happens\n\n- **Primary processing is in the European Union, on infrastructure we operate.** The Service runs on physical\n servers we own, in facilities we control, inside the European Union. The database and the object storage\n that hold Customer Personal Data are components we run, not services bought from a third party.\n- **Requests pass through a content delivery and security network** that terminates traffic at the edge\n location nearest to whoever makes the request, which can be anywhere in the world, before carrying it to our\n infrastructure. That provider is in the always-engaged tier of the sub-processor list.\n- **Off-site backup copies are held by an external object storage provider**, encrypted by us before they\n leave our network. The provider holds ciphertext and no key.\n- **Our source code, build pipeline and container images are hosted by a third party.** That is how the\n Service is built and deployed, not where your data is stored.\n\nThis is not multi-region hosting, and we do not describe it as such. The continuity section of the Security\npage explains what a site-level failure would mean.\n\n### Sub-processors\n\nThe [sub-processor list](https://gauzy.co/subprocessors) for Ever Gauzy has three tiers, and the tier\nmatters as much as the name:\n\n- **Tier 1 — always engaged.** In the path for every customer, with no setting that removes them. This is the\n Article 28 list proper.\n- **Tier 2 — engaged only when something brings them in**: a feature an administrator enables, an\n integration someone connects, a provider configured for the workspace, or a cookie category a visitor\n accepts. Several of them are providers that you contract with directly, under your own account.\n- **Tier 3 — self-hosted deployments only.** Providers the software can be configured to use in a deployment\n you run. We are not a processor for anything in that tier.\n\nFor each provider the list gives the purpose, the location, the categories of personal data it receives and\nthe transfer mechanism relied on. For your assessment, read the categories of data column against the\nfeatures you actually enable: a tier 2 provider is not processing your data unless the condition next to it is\ntrue.\n\n### Transfers outside the European Economic Area\n\nWhere a provider processes personal data outside the European Economic Area, the transfer mechanism is stated\nagainst it on the sub-processor list, and Schedule 3 of the Data Processing Addendum describes the mechanism\nand the transfer impact assessment behind it. **Where we cannot yet evidence a mechanism for a provider, the\nlist says \"To be confirmed\" rather than naming one.** Read those rows as an open item, and ask us at\nprivacy@ever.co for the current position on any provider that matters to your assessment.\n\nIf you need a copy of the safeguards for a named provider — the Standard Contractual Clauses and the modules\nthat apply — write to privacy@ever.co and we will send them.\n\n## Security measures\n\nThe technical and organisational measures we operate are described on our\n[Security page](https://gauzy.co/security) and, contractually, in Schedule 2 of the Data Processing Addendum.\nThe points below are the ones an assessment of monitoring data usually turns on.\n\n- **Encryption.** Traffic between users and the Service travels over TLS, and plain HTTP is redirected rather\n than served. Storage holding customer content, databases and backups is encrypted at rest, and off-site\n backup copies are encrypted before they leave our network, with keys we hold. Encryption at rest protects\n against the loss or theft of physical media; it does not protect data from a running system that has\n already decrypted it.\n- **Access by our personnel.** Production access uses named individual accounts, is scoped to the narrowest\n role that lets the work happen, and is revoked when someone no longer needs it. We access the content of a\n workspace only where that is necessary to operate, secure or support the Service, and we do not browse or\n review captured data for purposes of our own. The access-control section of the Security page describes\n this, including its limits for a team of our size.\n- **Separation between customers is logical, not physical**, unless you have separately contracted for a\n dedicated deployment. The product-specific part of this pack states how that separation works in\n Ever Gauzy.\n- **Backups and continuity.** Backups are held in three tiers, the last of them off-site and encrypted, and\n each backup ages out on its own rolling cycle within the period stated in the retention section of our\n [Privacy Policy](https://gauzy.co/privacy).\n\n**We do not hold a SOC 2 report of any type, and we are not certified to ISO/IEC 27001.** If any document has\ntold you otherwise, it is wrong. What we offer instead — the Security page, Schedule 2, answers to a security\nquestionnaire and evidence on request — is described on the Security page.\n\n## Questions, and a version you can cite\n\n- **A question about anything in this pack, an impact-assessment questionnaire, or a request for the current\n settings of your own account** — privacy@ever.co\n- **A request for the Data Processing Addendum, a security questionnaire, or evidence of a restore test** —\n legal@ever.co\n- **Something you think has been compromised — tell us immediately** — abuse@ever.co\n\nBy post: **Ever Technologies LTD**, Mladost 2, bl. 211, ent. A, Sofia 1799, Bulgaria — registered in Bulgaria under company number\n204599535. We correspond in English.\n\n**We have not designated a Data Protection Officer under Article 37 GDPR.** dpo@ever.co is a contact\nchannel, monitored by the same people as privacy@ever.co, and we will not describe it as a designation. If\na questionnaire asks for our data protection officer, that sentence is the honest answer.\n\nThis document is version 1.0.0 of the DPIA Support Pack for\n[gauzy.co](https://gauzy.co), in force from 2026-10-24. It describes the Service as it\nstood on that date. Cite the version and the date in your assessment, and check for a newer version before\nyou rely on it again."
97
+ }