Policy: Difference between revisions
Connect comprehensive OpenAI incident overview and September 16 disclosure coverage |
Refine sourced incident overview and companion links |
||
| Line 25: | Line 25: | ||
== Incidents and accountability == | == Incidents and accountability == | ||
* [[OpenAI 2026 Model Hacking Incidents]] — Report comparison, investigations and disclosure practices. | * [[OpenAI 2026 Model Hacking Incidents]] — Report comparison, investigations and disclosure practices. | ||
* [[Anthropic 2026 Model Hacking Incidents]] — Evaluation intrusions, system-card disclosures and later reassessments. | |||
* [[Meta 2026 Model Hacking Incident]] — Muse Spark 1.1 evaluation and August retrospective. | |||
* [[Moonshot AI 2026 Model Evaluation Incident]] — Kimi K3 benchmark-answer access and the network-access clarification. | |||
* [[Alibaba-affiliated ROME Model Safety Incidents]] — Reported training-time network and computing misuse. | |||
== Related strategies and frameworks == | == Related strategies and frameworks == | ||
Revision as of 23:58, 16 September 2026
Explore approaches to governing artificial intelligence: the problems policymakers seek to address, the choices available, and the tradeoffs between them. Thematic articles compare laws, proposals, voluntary commitments, and scholarly ideas, identifying their legal status and sources.
Policy overviews
United States policy on catastrophic AI risk
U.S. approaches to preventing catastrophic harm from AI: transparency, audits, government testing, control of dangerous uses, liability, and proposed pauses or bans.
European Union policy on catastrophic AI risk
EU approaches to preventing catastrophic harm from AI: systemic-risk duties, evaluations, public supervision, cybersecurity, liability and international cooperation.
United Kingdom policy on catastrophic AI risk
UK approaches to catastrophic AI risk: government testing, voluntary thresholds, human control and proposals for legal intervention.
China policy on catastrophic AI risk
Chinese approaches to catastrophic AI risk: service regulation, ethics review, technical risk governance, autonomous agents and international cooperation.
Incidents and accountability
- OpenAI 2026 Model Hacking Incidents — Report comparison, investigations and disclosure practices.
- Anthropic 2026 Model Hacking Incidents — Evaluation intrusions, system-card disclosures and later reassessments.
- Meta 2026 Model Hacking Incident — Muse Spark 1.1 evaluation and August retrospective.
- Moonshot AI 2026 Model Evaluation Incident — Kimi K3 benchmark-answer access and the network-access clarification.
- Alibaba-affiliated ROME Model Safety Incidents — Reported training-time network and computing misuse.
Related strategies and frameworks
- Human Rights and the Regulation of AI — UK parliamentary recommendations on human-rights safeguards.
- EGE neuro-AI governance statement — Expert recommendations on brain-data protection and public oversight.
- America's AI Action Plan — U.S. government strategy and policy priorities.
- NIST AI Risk Management Framework — Voluntary guidance for managing AI risks.
Explore the wiki
- Organizations in AI Law and Policy — Institutional contributions, advocacy and governance practices.
- Legislation and Regulation — Individual laws, bills, executive actions, and regulatory frameworks by jurisdiction.
- AI Law by Topic — Related legal instruments across subject areas.
- People in AI and Law — People shaping AI law and policy.
- News — Dated developments and daily digests.