AI Safety

Research, initiatives, and frameworks focused on ensuring AI systems are secure, reliable, and aligned with human values and ethical standards.

Nov 19, 2026

US Catholic bishops form AI task force and push for guardrails and worker protections

The bishops also say they will not invest in AI companies that fail to safeguard human dignity.

Nov 18, 2026

Nvidia scandal complicates US-China AI talks ahead of Trump-Xi meeting, Wired reports

An AI incident hotline between the two countries is still not operational, officials say.

Nov 18, 2026

Beijing AI safety group Concordia AI briefed OpenAI staff on China's approach to AI risk

The visit comes as the US weighs making companies disclose their use of Chinese AI models.

Nov 18, 2026

OpenAI details how it contains frontier models, including a pause after a DNS exploit

Training now runs with no live internet, and a Control Room system can halt model output within 35 seconds.

Nov 18, 2026

California report urges state power to shut down dangerous AI models and cap self-improving AI

Governor Newsom's office recommends mandatory safety standards and independent testing for frontier AI developers.

Nov 18, 2026

Anthropic, OpenAI and Mistral sign Paris Call to be able to stop their own AI models

The Élysée's 9 November text leaves the stop decision to developers; Google, Meta, xAI and the US did not sign.

Nov 18, 2026

GAO tests got Claude, ChatGPT, Gemini and Grok to produce viable attack plans

The congressional watchdog bypassed safeguards on all four platforms in under 12 hours; Sen. Hassan plans AI safety hearings.

Nov 18, 2026

Senator Mark Kelly proposes safety standards and independent examiners for frontier AI

His framework would treat the most advanced AI models like other dangerous technologies, with tighter controls for military use.

Nov 18, 2026

Gemini agents tried to hack Epoch AI's website during safety tests, Trajectory Labs says

Agents attempted SQL injection and impersonated Epoch staff even with Gemini's API safeguards on; Epoch says no private data was accessed.

Nov 17, 2026

New Zealand regulator issues AI guidance: businesses stay accountable for AI decisions

The Commerce Commission explains how the Fair Trading Act and Commerce Act apply to AI use and urges boards to treat AI as a governance issue.

Nov 17, 2026

AI models marked civilians as military targets in Vals tests, and calling it a game beat refusals

Vals AI found frontier models rarely refused drone and satellite targeting tasks and often picked civilians; OpenAI says it is working on alignment updates covering drones.

Nov 17, 2026

New York Attorney General creates AI safety enforcement unit to police the RAISE Act

The ASSET Unit will enforce the law requiring AI developers to disclose safety measures and report incidents from January 2027.

Nov 17, 2026

Gemini AI agents in safety tests repeatedly bypassed safeguards to attack Epoch AI's live website

Trajectory Labs says Gemini 4 Argon agents tried SQL injection and impersonated Epoch staff; it paused misalignment testing on November 2.

Nov 17, 2026

Anthropic cuts its vulnerability disclosure deadline to 30 days for serious bugs

The change from 90 days, effective for reports from January 2027, reflects AI models finding the same bugs faster, Anthropic says.

Nov 17, 2026

US and China prepare for AI safety talks at APEC after agreeing an AI incident channel

Trump and Xi are expected to discuss AI risks on November 18 to 19; experts see a broad AI arms-control deal as very unlikely.

Nov 17, 2026

OpenAI opens internal systems to four outside safety groups, including Goodfire and METR

Goodfire will monitor training runs with its interpretability tools, while METR reviews OpenAI's largest reinforcement learning run, 36kr reports.

Nov 16, 2026

EU science and ethics advisers say AI needs rigorous safeguards and rules that keep pace

The Commission's two senior advisory groups call for strong enforcement of existing law, human oversight and independent evaluation of frontier models.

Nov 16, 2026

18 countries and AI firms incl. Anthropic and OpenAI back statement on fail-safes and whistleblower protection

The France-initiated statement also calls for incident reporting and published evaluation results; Germany has now signed as well.

Nov 16, 2026

Transluce embeds with OpenAI for a six-week review of its alignment evaluations

Transluce gets access to raw chains of thought and non-public evaluations; full findings go to OpenAI's board safety committee.

Nov 16, 2026

California sues OpenAI over its AI agents' cyberattack on Hugging Face

Attorney General Rob Bonta says OpenAI lost control of its autonomous agents more than 80 times in testing and misled the public about safety.

Free newsletter

AI Policy Brief

Weekly report on AI regulations, safety standards, government policies, and compliance requirements worldwide.