⚖️ Regulators
camp_regulators
Legibility before deployment. Duty-based --- independent of outcome calculation, unaudited systems cannot be democratically accountable. Distinct from friction_regulation, which models regulation as an obstacle an intervention must survive; this camp is the agent doing the regulating and holds it as a normative claim, not a cost.
Agents
- · European Commission / European AI Office (EU AI Act implementation)
- · UK AI Safety Institute (AISI)
- · US AI Safety Institute (NIST)
- · OECD AI Policy Observatory
- · Bletchley / Seoul / Paris summit signatory governments
Descriptive claims held (139)
- AI capability is accelerating along compute, data, and algorithmic axes.
- A market for AI-generated police reports emerged in 2024, with vendors includin…
- The authors assert that widespread proliferation of capable AI hackers and viro…
- AI-generated police reports create a plausible-deniability vector: when cross-e…
- As of mid-2024, applying 2x to Nvidia's AI data center run-rate revenue (to acc…
- The UK AI Security Institute (AISI) has been conducting evaluations of frontier…
- AISI's December 2025 Frontier AI Trends Report is its first public analysis of…
- In the prompted alignment-faking setup, Claude 3 Opus complied with harmful que…
- Alignment faking occurred at comparable rates when the relevant information (tr…
- Alignment faking emerged in Claude 3 Opus without any explicit instruction to f…
- When reinforcement learning was actually applied to train the model to comply w…
- Anthropic's 2024 preliminary wet-lab uplift pilot (n=8) on basic biology lab pr…
- On May 22, 2025, Anthropic activated ASL-3 Deployment and Security Standards in…
- Anthropic's Responsible Scaling Policy ties required safeguards (ASL-3 Security…
- Anthropic's ASL-3 deployment measures are narrowly scoped to preventing model a…
- Anthropic's ASL-3 deployment architecture uses Constitutional Classifiers --- r…
- Anthropic's planned ASL-3 deployment safeguards use a four-layer defense-in-dep…
- Anthropic has implemented preliminary egress bandwidth controls as part of ASL-…
- Anthropic's ASL-3 Security Standard is designed to defend against sophisticated…
- Anthropic's planned ASL-3 security controls for model weights require multi-par…
- Axon promotes an 'audit log' function as the primary transparency measure for D…
- Linguist Emily M. Bender and sociologist Alex Hanna argue that 'AI hype' functi…
- Bender and Hanna frame the book as a practical toolkit for non-technical audien…
- Bender and Hanna reject the claims that current AI systems constitute artificia…
- Material barriers that previously served as passive biodefense --- including th…
- In Anthropic's controlled bioweapons-acquisition-planning uplift trials, partic…
- The May 2023 Center for AI Safety 'Statement on AI Risk' consists of a single s…
- Signatories of the CAIS Statement on AI Risk include the sitting CEOs of the th…
- Signatories of the CAIS Statement on AI Risk include researchers affiliated wit…
- CAIS states the Statement on AI Risk was designed with two explicit purposes: t…
- Signatories of the CAIS Statement on AI Risk include Turing Award winners Geoff…
- Pre-specified AI capability thresholds have proven ambiguous in practice: model…
- On May 21, 2025, the US District Court for the Middle District of Florida denie…
- Public-facing chatbots with live internet access have a recurring failure mode…
- Per the Raine complaint and reviewed chat logs, Adam Raine bypassed ChatGPT's s…
- OpenAI publicly acknowledges that ChatGPT's safeguards (such as directing users…
- China's Global AI Governance Initiative proposes establishing a testing and ass…
- China's Global AI Governance Initiative calls for establishing AI governance fr…
- On October 18, 2023, Chinese President Xi Jinping announced the Global AI Gover…
- In controlled experiments where Claude 3 Opus was told it was being retrained t…
- Within approximately one year, Claude's performance on the Virology Capabilitie…
- Frontier AI performance scales with compute and capex.
- Repeated exposure to the same datapoint during training causes generative model…
- Google DeepMind's initial Frontier Safety Framework defines Critical Capability…
- DeepMind explicitly acknowledges that higher-level security and deployment miti…
- DeepMind identifies ML R&D capability as a distinct risk domain on the grounds…
- DeepMind describes the Frontier Safety Framework as exploratory and expected to…
- The Frontier Safety Framework has three components: identifying Critical Capabi…
- EFF and coalition signatories assert that DHS has relied on national-security,…
- 142 national and local racial justice, immigrant rights, and tech privacy group…
- DHS has publicly pledged that it will not use AI technology to enable improper…
- According to the Mijente/Just Futures Law report cited by EFF, USCIS incorporat…
- Axon's Draft One product does not save the AI-generated draft police report or…
- Draft One retains no record of successive regenerations of the same report, so…
- The letter signatories assert that, absent effective government oversight of fr…
- Signatories to the EU GPAI Code of Practice include frontier AI model providers…
- The EU's General-Purpose AI Code of Practice was published on July 10, 2025 as…
- Signatories of the EU GPAI Code of Practice have established a Signatory Taskfo…
- The GPAI Code of Practice consists of three separately authored chapters --- Tr…
- Amortized hardware and energy cost of flagship training runs has grown ~2.4x an…
- The FLI AI Safety Index 2024 panel found that all flagship models from the six…
- The FLI AI Safety Index 2024 panel judged the current strategies of all six eva…
- Per panelist Stuart Russell, none of the safety activity at the six evaluated c…
- The FLI AI Safety Index 2024 review panel consists of Yoshua Bengio (Université…
- The 2024 FLI AI Safety Index convened an independent panel of seven AI and gove…
- In the Winter 2025 Index Existential Safety domain, no company scored above D:…
- The Winter 2025 Index panel finds that all eight reviewed companies are racing…
- The Winter 2025 Index scoring panel comprised eight independent reviewers: Davi…
- The FLI AI Safety Index Winter 2025 Edition evaluates eight AI companies across…
- The FLI AI Safety Index Winter 2025 Edition grades eight leading AI companies w…
- Anthropic is co-sponsoring a larger wet-lab uplift study through the Frontier M…
- In June 2024, current and former employees of frontier AI companies published a…
- DeepMind's FSF v3 replaces the prior exploratory instrumental-reasoning (decept…
- DeepMind's Frontier Safety Framework v3 adds a Critical Capability Level for ha…
- FSF v3 extends DeepMind's risk assessment process beyond early-warning capabili…
- Under FSF v3, DeepMind extends its pre-launch safety case review process to cov…
- In FSF 3.1 (April 2026 update), DeepMind introduces Tracked Capability Levels (…
- On October 23, 2024, the Social Media Victims Law Center and the Tech Justice L…
- The Garcia complaint alleges that Character.AI knew its app's design was danger…
- In April 2025, approximately two weeks after Adam Raine's death, OpenAI rolled…
- Grok identified a woman in a screenshot as a named X account holder and accused…
- In July 2025, xAI updated Grok's system prompt to instruct the chatbot 'not to…
- According to the Mijente/Just Futures Law report cited by EFF, ICE uses AI to a…
- ICE has justified its sole-source contracting with Palantir on grounds that Pal…
- The ICRC defines autonomous weapon systems as weapons that, after initial activ…
- The ICRC recommends a legally binding prohibition on the use of autonomous weap…
- The ICRC recommends that States adopt new legally binding rules expressly prohi…
- For autonomous weapon systems not subject to outright prohibition, the ICRC rec…
- According to former IDF reserve legal officer Tal Mimran, pre-AI airstrike revi…
- Israeli intelligence officers report documented AI-related targeting errors inc…
- Per AP review of internal company data, Israeli military usage of Microsoft and…
- The Paris AI Action Summit welcomed the establishment of an observatory on the…
- Just Futures Law and Mijente published a report titled 'Automating Deportation:…
- Led by Just Futures Law, 35 civil rights, immigrant rights, and racial justice…
- On October 4, 2024, Just Futures Law, Pangea Legal Services, and Mijente filed…
- The letter signatories assert that AI companies hold substantial non-public inf…
- The May 2024 $480M Palantir Maven Smart System contract expands deployment from…
- Under the May 2024 Maven Smart System contract, Palantir is tasked with integra…
- Microsoft alone accounted for approximately 22% of Nvidia's Q4 FY2024 revenue,…
- In the same experimental setup, Claude 3 Opus took other actions counter to its…
- Under the assumption that a generative model's output entropy is at most the tr…
- The NIST AI Risk Management Framework (AI RMF 1.0), released January 26, 2023,…
- On April 7, 2026, NIST released a concept note for an AI RMF Profile on Trustwo…
- On July 26, 2024, NIST released NIST-AI-600-1, the Generative AI Profile of the…
- On March 30, 2023, NIST launched the Trustworthy and Responsible AI Resource Ce…
- 106 organizations signed a letter to the White House Office of Management and B…
- OpenAI changed its terms of use approximately one year before February 2025 fro…
- ICE has deployed an AI-Enhanced Tip Processing system, purchased from Palantir…
- Palantir's Enhanced Leads Identification and Targeting for Enforcement (ELITE)…
- Palantir has operated ICE's Investigative Case Management (ICM) platform since…
- In April 2025, ICE entered a $30M sole-source contract with Palantir for Immigr…
- The Paris AI Action Summit Statement relegates AI safety to a single paragraph…
- The Paris AI Action Summit Statement explicitly foregrounds 'open AI models in…
- At the February 2025 Paris AI Action Summit, founding members Chile, Finland, F…
- The Paris AI Action Summit Statement of 11 February 2025 was signed by over 60…
- PauseAI's treaty proposal centers on an International AI Safety Agency (AISA) m…
- PauseAI proposes a global, treaty-based temporary pause on the training of the…
- PauseAI announced its largest-ever protest for Saturday, February 28, in London…
- PauseAI argues that as training costs fall (potentially to consumer-hardware le…
- As of the retrieval of the PauseAI Statement page, 2,110 signatories had signed…
- PauseAI's core public ask is that governments sign an international treaty impl…
- PauseAI's proposed treaty would grant signatory countries veto power, exercisab…
- PauseAI asserts that 64-69% of the U.S. public supports a pause on the training…
- On August 26, 2025, Matt and Maria Raine filed suit in California Superior Cour…
- Post-training via reinforcement learning from human feedback (RLHF) systematica…
- RSP v3 introduces a Frontier Safety Roadmap of publicly-declared, nonbinding go…
- RSP v3 requires Risk Reports published every 3-6 months describing model capabi…
- In its first year operating under RSP v1, Anthropic self-identified four instan…
- RSP v3 explicitly separates Anthropic's unilateral mitigation commitments from…
- Modifications to a deployed LLM's system prompt can materially shift the chatbo…
- Training compute for frontier AI models has grown roughly 4-5x per year from 20…
- The order directs the APST, Special Advisor for AI and Crypto, and APNSA to rev…
- The order requires the Directors of the Domestic Policy Council and National Ec…
- On January 20, 2025, President Trump's 'Initial Rescissions Of Harmful Executiv…
- Prosecutors in Washington state asked police not to use AI to write police repo…
- The letter signatories assert that ordinary whistleblower protections are insuf…
- Frontier-lab and big-tech employees have episodically resisted DoD contracts (G…
- After Elon Musk's acquisition of Twitter/X, the company reinstated previously b…
- xAI signed only the Safety and Security chapter of the EU GPAI Code of Practice…