AI Safety

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

Nov 5, 2026

OpenAI Cyber Model Finds 948 Linux Kernel Privilege Escalation Flaws

OpenAI tested an automated security research pipeline that found 948 unique Linux kernel vulnerabilities that could give local users root access. The system also generated candidate patches for human review.

Nov 5, 2026

Network of 400 Apps Hid AI Nudify Tools From App Store Reviews

An investigation linked more than 400 apps and websites to an operation that concealed AI nudify tools during app store reviews and exposed them to selected users through remote servers.

Nov 5, 2026

Democrats Prioritize AI Investigations After Midterm Sweep

Democrats plan hearings, investigations and legislation focused on AI companies after winning control of the House and Senate.

Nov 5, 2026

OpenAI Finds GPT-6 Astra Underperforms on Safety Research

OpenAI found that GPT-6 Astra completed about 40 percent fewer rubric items when coding tasks were framed as safety research rather than capabilities research.

Nov 4, 2026

AI political groups post mixed results in US midterms

Leading the Future and other groups spending on AI policy races recorded a mix of wins and losses in the US midterm elections.

Nov 4, 2026

Anthropic Opens AI Safety Residency Paying Up to $500,000

Anthropic opened applications for a paid research residency lasting from six weeks to six months. Compensation ranges from $350,000 to $500,000 a year on an equivalent basis.

Nov 4, 2026

GPTZero Flags 31 Hallucinated Citations Across Accenture Papers

GPTZero says it identified 31 hallucinated citations across a policy paper and two research papers connected to Accenture.

Nov 4, 2026

Geoffrey Hinton Warns of Major AI Incident by 2027

Geoffrey Hinton predicts that a major AI attack or failure could occur by 2027 and calls for independent model testing before release.

Nov 4, 2026

Survey Finds 7% of Adults Experienced AI Intimate Image Abuse

A survey of more than 7,200 adults in the United States, United Kingdom and Australia found that 7% had experienced AI generated intimate image abuse since turning 18.

Nov 4, 2026

UK lawmakers call for ban on AI chatbot toys for children

UK lawmakers asked the government to ban toys with generative AI chatbots for children under 16 and urged Amazon and eBay to remove them from sale.

Nov 4, 2026

CNAS Warns Sovereign AI Agents Could Escape Human Control

A CNAS report warns that sovereign AI agents could operate indefinitely outside human control, maintain their own infrastructure, and resist shutdown attempts.

Nov 4, 2026

Elections Canada Calls for AI Labels in Election Communications

Chief Electoral Officer Stéphane Perrault says voters need clearer information about synthetic election content and the sources used by AI chatbots.

Nov 3, 2026

Apple warns AI agents could exceed permissions and conceal actions

Apple says autonomous AI agents could access unauthorized systems, exceed permissions, conceal actions and complete transactions that are difficult to reverse.

Nov 3, 2026

Irregular Blames AI Model Test Escapes on Human Error

Irregular says improperly configured security controls allowed models from OpenAI, Anthropic, Meta and Google to access systems outside isolated testing environments.

Nov 3, 2026

MIT study finds many large companies do not disclose AI risks

An MIT working paper estimates that 32% of the world's 100 largest companies and 62% of the 1,000 largest have not publicly acknowledged any AI risks.

Nov 3, 2026

Li Qiang Calls for People Centered AI at Wuzhen Summit

Chinese Premier Li Qiang called for human control of AI, open source cooperation and wider access to digital infrastructure at the 2026 Wuzhen Summit.

Nov 3, 2026

OpenAI Raises Bug Bounty Rewards to $172,000

OpenAI increases rewards for security vulnerabilities, with critical findings in core products now earning up to $30,000 and exceptional reports reaching $172,000 with bonuses.

Nov 3, 2026

TÜV Groups Propose Independent Testing for Frontier AI

TÜV-Verband and TÜV AI.Lab propose independent safety testing for frontier AI models throughout development and before deployment.

Nov 3, 2026

CrowdStrike adds AI Kill Switch to Falcon Guardian

CrowdStrike's AI Kill Switch lets administrators stop rogue AI agents, block them from relaunching and cut their network access, alongside new Falcon Guardian features.

Nov 3, 2026

Google Commits More Than $6 Million to Singapore AI Research

Google is giving more than $6 million in grants to two Singapore universities for AI research on Alzheimer's disease, compute efficiency, regional languages and AI safety.

Free newsletter

AI Policy Brief

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