-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy path4.1_responsibility_part1.html
More file actions
571 lines (487 loc) · 39 KB
/
4.1_responsibility_part1.html
File metadata and controls
571 lines (487 loc) · 39 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>4.1 Disinformation, surveillance, and influence at scale - Responsibility</title>
<link href="https://fonts.googleapis.com/css2?family=Figtree:wght@300;400;500;600;700&display=swap" rel="stylesheet">
<style>
* {
margin: 0;
padding: 0;
box-sizing: border-box;
}
body {
font-family: 'Figtree', -apple-system, BlinkMacSystemFont, 'Segoe UI', sans-serif;
background-color: #ffffff;
color: #000000;
line-height: 1.3;
}
.container {
max-width: 1200px;
margin: 0 auto;
padding: 8px;
flex: 1;
min-width: 200px;
overflow-wrap: break-word;
word-break: break-word; }
h1 {
text-align: center;
margin-bottom: 8px;
color: #000000;
font-weight: 600;
font-size: 18px;
}
.legend {
text-align: center;
font-size: 12px;
color: #888888;
font-style: italic;
margin-bottom: 12px;
padding: 8px;
background-color: #f9f9f9;
border-radius: 5px;
border: 1px solid #e0e0e0;
}
.selection-title {
text-align: center;
font-size: 14px;
font-weight: 600;
color: #666666;
margin-bottom: 10px;
}
.nav-pills {
display: flex;
flex-wrap: wrap;
gap: 4px;
margin-bottom: 15px;
justify-content: center;
}
.nav-pill {
background: #f8f9fa;
border: 1px solid #e0e0e0;
border-radius: 25px;
padding: 12px 20px;
cursor: pointer;
font-family: 'Figtree', sans-serif;
font-size: 16px;
font-weight: 500;
transition: all 0.3s ease;
color: #000000;
}
.nav-pill:hover {
background: #e9ecef;
border-color: #000000;
}
.nav-pill.active {
background: #000000;
color: white;
border-color: #000000;
}
.actor-section {
display: none;
}
.actor-section.active {
display: block;
}
.content-grid {
display: flex;
width: 100%;
gap: 4px;
}
.content-column {
background: #ffffff;
border: 1px solid #e0e0e0;
border-radius: 8px;
padding: 8px;
flex: 1;
min-width: 200px;
overflow-wrap: break-word;
word-break: break-word; }
.criteria-header {
font-size: 12px;
font-weight: 600;
margin-bottom: 15px;
padding-bottom: 10px;
border-bottom: 2px solid;
}
.criteria-header.higher {
color: #FF0000;
border-bottom-color: #FF0000;
}
.criteria-header.lower {
color: #2E5C8A;
border-bottom-color: #2E5C8A;
}
.summary-section {
margin-bottom: 20px;
}
.summary-text {
margin-bottom: 15px;
font-weight: 500;
color: #000000;
font-size: 15px;
}
.quote-details {
margin-top: 15px;
}
.quote-toggle {
cursor: pointer;
color: #000000;
font-weight: 500;
font-size: 16px;
background-color: #ffff00;
padding: 10px 15px;
border-radius: 4px;
display: inline-block;
}
.quote-toggle:hover {
color: #333333;
}
.quote-list {
margin-top: 15px;
padding-left: 20px;
}
.quote-list li {
margin-bottom: 12px;
font-size: 16px;
padding: 10px 15px;
line-height: 1.3;
color: #000000;
}
@media (max-width: 768px) {
.content-grid {
gap: 4px;
}
.selection-title {
text-align: center;
font-size: 14px;
font-weight: 600;
color: #666666;
margin-bottom: 10px;
}
.nav-pills {
justify-content: flex-start;
}
.nav-pill {
font-size: 16px;
padding: 4px 8px;
}
}
</style>
</head>
<body>
<div class="container">
<h1>4.1 Disinformation, surveillance, and influence at scale - Responsibility</h1>
<div class="selection-title">Select an actor:</div>
<div class="nav-pills">
<button class="nav-pill active" data-target="AIDeveloperGeneralpurposeAI">
AI Developer (General-purpose AI)
</button>
<button class="nav-pill" data-target="AIDeployer">
AI Deployer
</button>
<button class="nav-pill" data-target="AIGovernanceActor">
AI Governance Actor
</button>
<button class="nav-pill" data-target="AIUser">
AI User
</button>
</div>
<div class="content-sections">
<div class="actor-section active" id="AIDeveloperGeneralpurposeAI">
<div class="content-grid">
<div class="content-column">
<h3 class="criteria-header higher">Reasons for Higher Responsibility</h3>
<div class="summary-section">
<p class="summary-text"><strong>AI-generated summary:</strong> One expert commented: "Responsibility follows control over generation, distribution, and targeting rails. GP devs + infra are high as they shape capability surfaces, provenance, rate limits, and abuse tooling."</p>
<details class="quote-details">
<summary class="quote-toggle">See all expert comments (1)</summary>
<ul class="quote-list">
<li>"Responsibility follows control over generation, distribution, and targeting rails. GP devs + infra are high as they shape capability surfaces, provenance, rate limits, and abuse tooling."</li>
</ul>
</details>
</div>
</div>
<div class="content-column">
<h3 class="criteria-header lower">Reasons for Lower Responsibility</h3>
<div class="summary-section">
<p class="summary-text"><strong>AI-generated summary:</strong> [NO EXPERT COMMENTS PROVIDED]</p>
</div>
</div>
</div>
</div>
<div class="actor-section" id="AIDeployer">
<div class="content-grid">
<div class="content-column">
<h3 class="criteria-header higher">Reasons for Higher Responsibility</h3>
<div class="summary-section">
<p class="summary-text"><strong>AI-generated summary:</strong> Deployers bear primary to high responsibility because they run the critical control points for scaling disinformation: ranking algorithms, advertising systems, recommender knobs, API access, and content takedown decisions.</p>
<details class="quote-details">
<summary class="quote-toggle">See all expert comments (2)</summary>
<ul class="quote-list">
<li>"Upstream data providers have a particularly strong responsibility here. If a foundation model producer has outsourced their data aquisition to offshore providers, they may also have introduced sleeper agents into their models without realising. Anthropic's research on this topic showed Deployers and downstream developers have no realistic way to fine tune out sleeper agent behaviour, but that doesn't mean they shouldn't be trying to detect it and intervene."</li> <li>"Responsibility follows control over generation, distribution, and targeting rails. Deployers are primary as they run ranking, ads, recommender knobs, API access, and takedowns."</li>
</ul>
</details>
</div>
</div>
<div class="content-column">
<h3 class="criteria-header lower">Reasons for Lower Responsibility</h3>
<div class="summary-section">
<p class="summary-text"><strong>AI-generated summary:</strong> One expert commented: "AI Deployer: Quite confused on the consensus here. How exactly are those deploying responsible? What measures could they take to effectively combat the deployment of the model facilitating misinformation? Perhaps for surveillance they carry some further degree of responsibility, but otherwise it seems like responsibility would naturally fall on the developer (who could set sufficient safeguards) or governance actor (as this risk will often play out at scale, implicating a diffuse responsibility that will likely need coordination to address). "</p>
<details class="quote-details">
<summary class="quote-toggle">See all expert comments (1)</summary>
<ul class="quote-list">
<li>"AI Deployer: Quite confused on the consensus here. How exactly are those deploying responsible? What measures could they take to effectively combat the deployment of the model facilitating misinformation? Perhaps for surveillance they carry some further degree of responsibility, but otherwise it seems like responsibility would naturally fall on the developer (who could set sufficient safeguards) or governance actor (as this risk will often play out at scale, implicating a diffuse responsibility that will likely need coordination to address)."</li>
</ul>
</details>
</div>
</div>
</div>
</div>
<div class="actor-section" id="AIGovernanceActor">
<div class="content-grid">
<div class="content-column">
<h3 class="criteria-header higher">Reasons for Higher Responsibility</h3>
<div class="summary-section">
<p class="summary-text"><strong>AI-generated summary:</strong> Governance actors bear high to extreme responsibility—even more than private companies—because it ultimately falls on them to protect the public and enact regulations and enforcement. They set rules and oversight mechanisms that prevent or enable misuse, have obligation to regulate and capability to enforce standards, and must mandate provenance/labeling, acceptable use policy enforcement, and cross-platform response coordination. Intelligence agencies may need to retain records of citizen-AI contacts to understand foreign control mechanisms, especially regarding election influence operations.</p>
<details class="quote-details">
<summary class="quote-toggle">See all expert comments (4)</summary>
<ul class="quote-list">
<li>"Upstream data providers have a particularly strong responsibility here. If a foundation model producer has outsourced their data aquisition to offshore providers, they may also have introduced sleeper agents into their models without realising. Anthropic's research on this topic showed Deployers and downstream developers have no realistic way to fine tune out sleeper agent behaviour, but that doesn't mean they shouldn't be trying to detect it and intervene.
Ultimately intelligence agencies may need to retain records of all contacts between citizens and AIs if they wish to understand foreign ownership and control mechanisms within their society, as it may be less transparently interogatable than other telecommunications techniques as a means of hiding correspondence and influence.
This is particularly pronounced when countries are trying to influence each other's elections. In Taiwan the CCP created AI video of ballot boxed being stuffed. In Ukraine russia has used AI to generate memes and social media virality. The US is particularly adept at these techniques and as the major model provider it should be expected that it will leverage it as a soft power and foreign influence tool, if not as a recruitment and asset handling service."</li> <li>"Developers and deployers bear primary responsibility given their direct control over system design and deployment choices. Governance actors are crucial for setting standards and enforcement. Users and affected stakeholders have limited responsibility but should practice informed usage."</li> <li>"I think governance actors have extreme responsibility for managing these problems, even more than private companies. It ultimately falls on them to protect the public and enact regulations and enforcement."</li> <li>"Responsibility follows control over generation, distribution, and targeting rails. Governance is high to mandate provenance/labeling, AUP enforcement, and cross-platform response."</li>
</ul>
</details>
</div>
</div>
<div class="content-column">
<h3 class="criteria-header lower">Reasons for Lower Responsibility</h3>
<div class="summary-section">
<p class="summary-text"><strong>AI-generated summary:</strong> One expert commented: "I continue to assess that AI governance actors are "minimally responsible". I think there is a category error at work here. Saying AI governance actors are responsible for AI risks is like saying that judges are responsible for crimes being committed. The kind of responsibility a judge has is very different from the kind of responsibility that a criminal or a lock-pick-maker has. The better way to think of this is that the AI governance actor is responsible for holding responsible the actor who is properly responsible. It would be recursive that AI governance actor is themselves responsible. Would we propose some meta-AI-goverance governor who holds responsible the AI governance actors that fail to hold responsible the actors that should be responsible? This is not the right way of thinking. We can rightly say that AI governance actors are responsible for some meta issues, like ensuring that Governments are properly informed etc."</p>
<details class="quote-details">
<summary class="quote-toggle">See all expert comments (1)</summary>
<ul class="quote-list">
<li>"I continue to assess that AI governance actors are "minimally responsible". I think there is a category error at work here. Saying AI governance actors are responsible for AI risks is like saying that judges are responsible for crimes being committed. The kind of responsibility a judge has is very different from the kind of responsibility that a criminal or a lock-pick-maker has. The better way to think of this is that the AI governance actor is responsible for holding responsible the actor who is properly responsible. It would be recursive that AI governance actor is themselves responsible. Would we propose some meta-AI-goverance governor who holds responsible the AI governance actors that fail to hold responsible the actors that should be responsible? This is not the right way of thinking. We can rightly say that AI governance actors are responsible for some meta issues, like ensuring that Governments are properly informed etc."</li>
</ul>
</details>
</div>
</div>
</div>
</div>
<div class="actor-section" id="AIUser">
<div class="content-grid">
<div class="content-column">
<h3 class="criteria-header higher">Reasons for Higher Responsibility</h3>
<div class="summary-section">
<p class="summary-text"><strong>AI-generated summary:</strong> One expert commented: "I've ranked AI user higher than average, since users may include independent actors jailbreaking or fine-tuning models to run large-scale disinformation operations. One of the main concerns for AI-powered disinformation that it lowers the bar for entry for AI users to conduct large-scale ops. "</p>
<details class="quote-details">
<summary class="quote-toggle">See all expert comments (1)</summary>
<ul class="quote-list">
<li>"I've ranked AI user higher than average, since users may include independent actors jailbreaking or fine-tuning models to run large-scale disinformation operations. One of the main concerns for AI-powered disinformation that it lowers the bar for entry for AI users to conduct large-scale ops."</li>
</ul>
</details>
</div>
</div>
<div class="content-column">
<h3 class="criteria-header lower">Reasons for Lower Responsibility</h3>
<div class="summary-section">
<p class="summary-text"><strong>AI-generated summary:</strong> While they can practice informed usage, their responsibility should be limited given their constraints. One expert advocates for fairly modest responsibility assignment to regular users.</p>
<details class="quote-details">
<summary class="quote-toggle">See all expert comments (3)</summary>
<ul class="quote-list">
<li>"Developers and deployers bear primary responsibility given their direct control over system design and deployment choices. Governance actors are crucial for setting standards and enforcement. Users and affected stakeholders have limited responsibility but should practice informed usage."</li> <li>"I assign fairly modest responsibility for regular users who promote misinformation. Hard for them to do much better."</li> <li>"Responsibility follows control over generation, distribution, and targeting rails. Users are minimal; affected stakeholders none as they bear impacts but don't control the rails."</li>
</ul>
</details>
</div>
</div>
</div>
</div>
</div>
</div>
</div>
</div>
<div class="actor-section" id="AIDeveloperSpecializedAI">
<div class="content-grid">
<div class="content-column">
<h3 class="criteria-header higher">Reasons for Higher Responsibility</h3>
<div class="summary-section">
<p class="summary-text"><strong>AI-generated summary:</strong> Specialized developers bear high to primary responsibility because they create products like ad-tech, psychographics, and surveillance/OSINT tools that directly enable persuasion and monitoring at scale. Upstream data providers have particularly strong responsibility—if foundation model producers outsource data acquisition to offshore providers, they may introduce sleeper agents into models without realizing it, and research shows deployers and downstream developers have no realistic way to fine-tune out this behavior.</p>
<details class="quote-details">
<summary class="quote-toggle">See all expert comments (2)</summary>
<ul class="quote-list">
<li>"Upstream data providers have a particularly strong responsibility here. If a foundation model producer has outsourced their data aquisition to offshore providers, they may also have introduced sleeper agents into their models without realising. Anthropic's research on this topic showed Deployers and downstream developers have no realistic way to fine tune out sleeper agent behaviour, but that doesn't mean they shouldn't be trying to detect it and intervene."</li> <li>"Responsibility follows control over generation, distribution, and targeting rails. Specialized devs are primary when their products (ad-tech, psychographics, surveillance/OSINT) directly enable persuasion/monitoring."</li>
</ul>
</details>
</div>
</div>
<div class="content-column">
<h3 class="criteria-header lower">Reasons for Lower Responsibility</h3>
<div class="summary-section">
<p class="summary-text"><strong>AI-generated summary:</strong> [NO EXPERT COMMENTS PROVIDED]</p>
</div>
</div>
</div>
</div>
<div class="actor-section" id="AIDeployer">
<div class="content-grid">
<div class="content-column">
<h3 class="criteria-header higher">Reasons for Higher Responsibility</h3>
<div class="summary-section">
<p class="summary-text"><strong>AI-generated summary:</strong> Deployers bear primary to high responsibility because they run the critical control points for scaling disinformation: ranking algorithms, advertising systems, recommender knobs, API access, and content takedown decisions.</p>
<details class="quote-details">
<summary class="quote-toggle">See all expert comments (2)</summary>
<ul class="quote-list">
<li>"Upstream data providers have a particularly strong responsibility here. If a foundation model producer has outsourced their data aquisition to offshore providers, they may also have introduced sleeper agents into their models without realising. Anthropic's research on this topic showed Deployers and downstream developers have no realistic way to fine tune out sleeper agent behaviour, but that doesn't mean they shouldn't be trying to detect it and intervene."</li> <li>"Responsibility follows control over generation, distribution, and targeting rails. Deployers are primary as they run ranking, ads, recommender knobs, API access, and takedowns."</li>
</ul>
</details>
</div>
</div>
<div class="content-column">
<h3 class="criteria-header lower">Reasons for Lower Responsibility</h3>
<div class="summary-section">
<p class="summary-text"><strong>AI-generated summary:</strong> One expert commented: "AI Deployer: Quite confused on the consensus here. How exactly are those deploying responsible? What measures could they take to effectively combat the deployment of the model facilitating misinformation? Perhaps for surveillance they carry some further degree of responsibility, but otherwise it seems like responsibility would naturally fall on the developer (who could set sufficient safeguards) or governance actor (as this risk will often play out at scale, implicating a diffuse responsibility that will likely need coordination to address). "</p>
<details class="quote-details">
<summary class="quote-toggle">See all expert comments (1)</summary>
<ul class="quote-list">
<li>"AI Deployer: Quite confused on the consensus here. How exactly are those deploying responsible? What measures could they take to effectively combat the deployment of the model facilitating misinformation? Perhaps for surveillance they carry some further degree of responsibility, but otherwise it seems like responsibility would naturally fall on the developer (who could set sufficient safeguards) or governance actor (as this risk will often play out at scale, implicating a diffuse responsibility that will likely need coordination to address)."</li>
</ul>
</details>
</div>
</div>
</div>
</div>
<div class="actor-section" id="AIInfrastructureProvider">
<div class="content-grid">
<div class="content-column">
<h3 class="criteria-header higher">Reasons for Higher Responsibility</h3>
<div class="summary-section">
<p class="summary-text"><strong>AI-generated summary:</strong> Responsibility follows control over generation, distribution, and targeting rails. They wouldn't be granted safe harbor protections if harboring organized crime or foreign state actors, and have similar responsibility when models they host act as proxies for threat actors. They must detect and remove such intrusions, supply provenance and rate limit controls, and can facilitate or block mass disinformation and surveillance through their platforms.</p>
<details class="quote-details">
<summary class="quote-toggle">See all expert comments (2)</summary>
<ul class="quote-list">
<li>"I note others give lower responsibility to infrastructure providers. These entities would not be granted safe harbor protections if they were harboring organised crime or foreign state actors, they would have a responsibility to detect and remove such intrusions. Should the models they host be acting as a proxy for threat actors, they have a similar responsibility over the AI they're serving.
Upstream data providers have a particularly strong responsibility here. If a foundation model producer has outsourced their data aquisition to offshore providers, they may also have introduced sleeper agents into their models without realising."</li> <li>"Responsibility follows control over generation, distribution, and targeting rails. GP devs + infra are high as they shape capability surfaces, provenance, rate limits, and abuse tooling."</li>
</ul>
</details>
</div>
</div>
<div class="content-column">
<h3 class="criteria-header lower">Reasons for Lower Responsibility</h3>
<div class="summary-section">
<p class="summary-text"><strong>AI-generated summary:</strong> [NO EXPERT COMMENTS PROVIDED]</p>
</div>
</div>
</div>
</div>
<div class="actor-section" id="AIUser">
<div class="content-grid">
<div class="content-column">
<h3 class="criteria-header higher">Reasons for Higher Responsibility</h3>
<div class="summary-section">
<p class="summary-text"><strong>AI-generated summary:</strong> One expert commented: "I've ranked AI user higher than average, since users may include independent actors jailbreaking or fine-tuning models to run large-scale disinformation operations. One of the main concerns for AI-powered disinformation that it lowers the bar for entry for AI users to conduct large-scale ops. "</p>
<details class="quote-details">
<summary class="quote-toggle">See all expert comments (1)</summary>
<ul class="quote-list">
<li>"I've ranked AI user higher than average, since users may include independent actors jailbreaking or fine-tuning models to run large-scale disinformation operations. One of the main concerns for AI-powered disinformation that it lowers the bar for entry for AI users to conduct large-scale ops."</li>
</ul>
</details>
</div>
</div>
<div class="content-column">
<h3 class="criteria-header lower">Reasons for Lower Responsibility</h3>
<div class="summary-section">
<p class="summary-text"><strong>AI-generated summary:</strong> While they can practice informed usage, their responsibility should be limited given their constraints. One expert advocates for fairly modest responsibility assignment to regular users.</p>
<details class="quote-details">
<summary class="quote-toggle">See all expert comments (3)</summary>
<ul class="quote-list">
<li>"Developers and deployers bear primary responsibility given their direct control over system design and deployment choices. Governance actors are crucial for setting standards and enforcement. Users and affected stakeholders have limited responsibility but should practice informed usage."</li> <li>"I assign fairly modest responsibility for regular users who promote misinformation. Hard for them to do much better."</li> <li>"Responsibility follows control over generation, distribution, and targeting rails. Users are minimal; affected stakeholders none as they bear impacts but don't control the rails."</li>
</ul>
</details>
</div>
</div>
</div>
</div>
<div class="actor-section" id="AffectedStakeholder">
<div class="content-grid">
<div class="content-column">
<h3 class="criteria-header higher">Reasons for Higher Responsibility</h3>
<div class="summary-section">
<p class="summary-text"><strong>AI-generated summary:</strong> One expert commented: "Affected stakeholders also bear high responsibility - the spread of disinformation is so insidious because the messages are quickly picked up and amplified by independent communities, boosting the impact and value of a given disinformation op exponentially and organically. Thus the public has high obligation, capability and causal influence concerning the spread of disinformation. "</p>
<details class="quote-details">
<summary class="quote-toggle">See all expert comments (1)</summary>
<ul class="quote-list">
<li>"Affected stakeholders also bear high responsibility - the spread of disinformation is so insidious because the messages are quickly picked up and amplified by independent communities, boosting the impact and value of a given disinformation op exponentially and organically. Thus the public has high obligation, capability and causal influence concerning the spread of disinformation."</li>
</ul>
</details>
</div>
</div>
<div class="content-column">
<h3 class="criteria-header lower">Reasons for Lower Responsibility</h3>
<div class="summary-section">
<p class="summary-text"><strong>AI-generated summary:</strong> Responsibility follows control over generation, distribution, and targeting rails. Users are minimal; affected stakeholders none as they bear impacts but don't control the rails.</p>
<details class="quote-details">
<summary class="quote-toggle">See all expert comments (2)</summary>
<ul class="quote-list">
<li>"Developers and deployers bear primary responsibility given their direct control over system design and deployment choices. Governance actors are crucial for setting standards and enforcement. Users and affected stakeholders have limited responsibility but should practice informed usage."</li> <li>"Responsibility follows control over generation, distribution, and targeting rails. Users are minimal; affected stakeholders none as they bear impacts but don't control the rails."</li>
</ul>
</details>
</div>
</div>
</div>
</div>
<div class="actor-section" id="AIGovernanceActor">
<div class="content-grid">
<div class="content-column">
<h3 class="criteria-header higher">Reasons for Higher Responsibility</h3>
<div class="summary-section">
<p class="summary-text"><strong>AI-generated summary:</strong> Governance actors bear high to extreme responsibility—even more than private companies—because it ultimately falls on them to protect the public and enact regulations and enforcement. They set rules and oversight mechanisms that prevent or enable misuse, have obligation to regulate and capability to enforce standards, and must mandate provenance/labeling, acceptable use policy enforcement, and cross-platform response coordination. Intelligence agencies may need to retain records of citizen-AI contacts to understand foreign control mechanisms, especially regarding election influence operations.</p>
<details class="quote-details">
<summary class="quote-toggle">See all expert comments (4)</summary>
<ul class="quote-list">
<li>"Upstream data providers have a particularly strong responsibility here. If a foundation model producer has outsourced their data aquisition to offshore providers, they may also have introduced sleeper agents into their models without realising. Anthropic's research on this topic showed Deployers and downstream developers have no realistic way to fine tune out sleeper agent behaviour, but that doesn't mean they shouldn't be trying to detect it and intervene.
Ultimately intelligence agencies may need to retain records of all contacts between citizens and AIs if they wish to understand foreign ownership and control mechanisms within their society, as it may be less transparently interogatable than other telecommunications techniques as a means of hiding correspondence and influence.
This is particularly pronounced when countries are trying to influence each other's elections. In Taiwan the CCP created AI video of ballot boxed being stuffed. In Ukraine russia has used AI to generate memes and social media virality. The US is particularly adept at these techniques and as the major model provider it should be expected that it will leverage it as a soft power and foreign influence tool, if not as a recruitment and asset handling service."</li> <li>"Developers and deployers bear primary responsibility given their direct control over system design and deployment choices. Governance actors are crucial for setting standards and enforcement. Users and affected stakeholders have limited responsibility but should practice informed usage."</li> <li>"I think governance actors have extreme responsibility for managing these problems, even more than private companies. It ultimately falls on them to protect the public and enact regulations and enforcement."</li> <li>"Responsibility follows control over generation, distribution, and targeting rails. Governance is high to mandate provenance/labeling, AUP enforcement, and cross-platform response."</li>
</ul>
</details>
</div>
</div>
<div class="content-column">
<h3 class="criteria-header lower">Reasons for Lower Responsibility</h3>
<div class="summary-section">
<p class="summary-text"><strong>AI-generated summary:</strong> One expert commented: "I continue to assess that AI governance actors are "minimally responsible". I think there is a category error at work here. Saying AI governance actors are responsible for AI risks is like saying that judges are responsible for crimes being committed. The kind of responsibility a judge has is very different from the kind of responsibility that a criminal or a lock-pick-maker has. The better way to think of this is that the AI governance actor is responsible for holding responsible the actor who is properly responsible. It would be recursive that AI governance actor is themselves responsible. Would we propose some meta-AI-goverance governor who holds responsible the AI governance actors that fail to hold responsible the actors that should be responsible? This is not the right way of thinking. We can rightly say that AI governance actors are responsible for some meta issues, like ensuring that Governments are properly informed etc."</p>
<details class="quote-details">
<summary class="quote-toggle">See all expert comments (1)</summary>
<ul class="quote-list">
<li>"I continue to assess that AI governance actors are "minimally responsible". I think there is a category error at work here. Saying AI governance actors are responsible for AI risks is like saying that judges are responsible for crimes being committed. The kind of responsibility a judge has is very different from the kind of responsibility that a criminal or a lock-pick-maker has. The better way to think of this is that the AI governance actor is responsible for holding responsible the actor who is properly responsible. It would be recursive that AI governance actor is themselves responsible. Would we propose some meta-AI-goverance governor who holds responsible the AI governance actors that fail to hold responsible the actors that should be responsible? This is not the right way of thinking. We can rightly say that AI governance actors are responsible for some meta issues, like ensuring that Governments are properly informed etc."</li>
</ul>
</details>
</div>
</div>
</div>
</div>
</div>
</div>
<script>
document.addEventListener('DOMContentLoaded', function() {
const pills = document.querySelectorAll('.nav-pill');
const sections = document.querySelectorAll('.actor-section');
pills.forEach(pill => {
pill.addEventListener('click', function() {
// Remove active class from all pills and sections
pills.forEach(p => p.classList.remove('active'));
sections.forEach(s => s.classList.remove('active'));
// Add active class to clicked pill
this.classList.add('active');
// Show corresponding section
const targetId = this.getAttribute('data-target');
const targetSection = document.getElementById(targetId);
if (targetSection) {
targetSection.classList.add('active');
}
});
});
});
</script>
</body>
</html>