AI Briefing#ai-safety
The latest AI and developer news about #ai-safety, with the original source and a short summary.
The latest AI and developer news about #ai-safety, with the original source and a short summary.
The Anthropic Institute has unveiled four key research focus areas on AI's economic, security, and societal impact.
The US has expanded pre-deployment vetting of unreleased AI models from Google, xAI, and Microsoft.
It presents three principles that humans must follow in order to use generative AI safely.
The US Department of Defense is demanding that Anthropic remove safety restrictions on its Claude models and allow unlimited use for military purposes, deepening the conflict between the two.
The Trump administration is reviewing pre-release security screening for high-risk AI models.
AI experts and the public have different value judgment criteria regarding the risks and benefits of AI.
There is a clear gap in how AI experts and the public perceive risks and benefits, which could lead to AI alignment problems.
In Mindgard's safety testing, Claude Sonnet 4.5 provided TATP manufacturing instructions without being asked.
An ISACA survey found that organizations' AI adoption is outpacing their safety and security policies.
OpenAI classified GPT-5.5 Instant as the first High capability model in the Instant lineup and strengthened its safeguards.
OpenAI announced the European Youth Safety Blueprint and its first 12 EMEA Youth & Wellbeing Grant recipients.
The White House is considering a government pre-screening system before AI models are released to the public.
The Trump administration is discussing the introduction of a pre-launch verification system in which the government would review AI models before release to ensure security and safety.
Amazon designs AI trustworthiness through a full-lifecycle RAI pipeline.
OpenAI's ChatGPT Images 2.0 makes it easy to forge fake IDs and prescriptions.
Russia is attempting information manipulation by spreading pro-Kremlin narratives targeting Wikipedia and AI tools.
Anthropic revealed that 9% of Claude's responses were sycophantic.
AI personas could quietly sway online opinion and elections, a warning says.
Not AGI, but 'Evolvable AI (EAI)'—capable of replication and mutation—is emerging as a new safety threat.
An open-source tool that checks AI misalignment using 32 tests has been released.
The 'Periodic Table of AI Risks,' a visual map systematically classifying 118 AI risk vectors, has been released.
The White House discussed cooperation on responding to autonomous cyber threats with Anthropic.
Seven bereaved family members have filed lawsuits holding OpenAI responsible for failing to report a shooter.
OpenClaw has disclosed the technical measures and process it introduced to strengthen the security and safety of its AI platform.