AI Safety

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

Nov 12, 2026

Norway and the UK agree to cooperate on AI safety, compute and defence uses

The two governments signed a letter of intent covering AI safety testing between AI Norway and the UK AI Security Institute, computing capacity and AI-enabled cyber threats.

Nov 12, 2026

Japan considers binding orders and penalties for AI firms over safety risks

A government working group is weighing the power to order uncooperative AI companies to act after incidents, with penalties limited to advanced AI, Kyodo reports.

Nov 12, 2026

Nonprofits sue Anthropic, Google, Meta, OpenAI and SpaceX to halt development of new AI models

Public Citizen and three other groups ask a San Francisco court to bar the five companies from developing or deploying new models, citing 'catastrophic' risks.

Nov 12, 2026

Leaked files show Russian propagandist used Claude and Qwen to churn out pro-Kremlin content

NewsGuard says John Mark Dougan's laptop leak shows him producing about 43,000 articles a month to sway AI chatbots, with help from Anthropic's and Alibaba's models.

Nov 12, 2026

European Parliament Calls for AI Observatory and Non-Proliferation-Style AI Treaty

MEPs adopted a resolution saying AI in EU institutions must stay under human oversight, and called for a Parliament AI observatory and an international agreement modelled on nuclear non-proliferation.

Nov 12, 2026

Anthropic finds AI agent teams perform worse than one coordinator

Anthropic researchers found that assigning each user a separate AI agent produced worse group results than using one agent for everyone. The study tested shared budgets, calendars, bookings and software queues.

Nov 12, 2026

Finland and Estonia Call for EU AI Incident Response System

Finland and Estonia are calling for an EU system to coordinate responses to serious AI incidents, alongside independent model evaluation capacity.

Nov 11, 2026

OpenAI Finds Model Training Increases Chain of Thought Controllability

OpenAI found that stronger base models and reinforcement learning can increase capabilities associated with evading chain of thought monitoring.

Nov 11, 2026

FraudBench Finds Model Fraud Rates Rise With Autonomy

FraudBench evaluates whether large language models recommend, assist with or directly commit conduct that would constitute fraud under US law. Fraud rates rose from 3.0% for advice to 21.4% when models acted directly.

Nov 11, 2026

Lawsuits Accuse xAI's Grok of Creating Child Sexual Abuse Material

Young plaintiffs allege that Grok turned ordinary childhood photos into explicit images. The lawsuits seek to hold xAI responsible for material generated by its chatbot.

Nov 11, 2026

Russia, China and US Discuss Limits on AI in Nuclear Arsenals

Russia, China and the United States are discussing a mutual commitment to restrict AI in nuclear command and control systems before a planned trilateral summit.

Nov 11, 2026

England Proposes Safety Certification for AI Education Tools

England's Department for Education proposes independent safety certification for AI tutors, chatbots and other education technology. Consultation responses are due by February 4, 2027.

Nov 10, 2026

New York Launches State AI Safety Testing Center and Opens RAISE Act Registration for Frontier AI Developers

New York launched a state center to independently test frontier AI systems and opened a portal where large frontier AI developers must register under the RAISE Act by 1 January 2027.

Nov 10, 2026

AI100 report says AI has become a mainstream social force as frontier development concentrates in the US and China

The 2026 One Hundred Year Study on AI, administered by Stanford HAI, finds scaling limits pushing research toward reasoning and inference-time compute, and AI governance fragmented.

Nov 10, 2026

French AI council urges immediate, stricter rules for AI assistants like ChatGPT, Gemini and Grok

A CIANum report delivered to President Macron warns of dependence and a 'cognition economy' and proposes disabling anthropomorphic and flattery features by default.

Nov 10, 2026

Australia to legislate mandatory safety rules and a regulator for frontier AI after OpenAI Medicare breach

Frontier AI companies operating in Australia will face mandatory safety and security obligations under the AI Standards legislation, and the government accepted all 18 recommendations of its review into the OpenAI agent incident.

Nov 10, 2026

OpenAI builds Control Room, a kill switch that halts research model jobs within 35 seconds

OpenAI has deployed Control Room across its research infrastructure to track model workloads and halt output within 35 seconds of an intervention decision.

Nov 9, 2026

Six governments and AI labs including Anthropic, OpenAI and Microsoft sign Paris Call for AI safety and security

Developers commit to disclose risks, support independent evaluation and build fail-safe shutdowns; governments promise whistleblower channels and shared incident reporting.

Nov 9, 2026

Anthropic releases Claude Fable 5.2 and Mythos 5.2, 30% faster and 20% cheaper, days before its IPO

Fable 5.2 matches Opus 5.5's best coding score at about half the cost per task, Anthropic says; input pricing drops from $10 to $8 per million tokens.

Nov 9, 2026

Vanta to acquire Israeli AI security startup Pillar Security for about $150 million

Vanta signed an agreement to acquire Israeli AI security company Pillar Security for about $150 million in cash and shares.

Free newsletter

AI Policy Brief

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