Updated daily503 verified stories
Safety News tracker.
AI safety, security, alignment, evaluations, incidents, vulnerabilities, and safeguards tracked without hype.
- 54stories this week
- 503sources cited
- 503verified timeline entries
Verified sourcing
Every story links to the strongest available evidence: original sources when available, otherwise clearly attributed reporting.
Plain English
What happened, why it matters, and what to watch — without the jargon.
No filler
When the signal is thin, we publish nothing rather than padding the feed.
More stories
9 storiesSecurity
Anthropic reports 190 million distillation attacks from Chinese AI labs
Anthropic disclosed that major Chinese AI labs, including Alibaba, Moonshot, and DeepSeek, launched nearly 190 million distillation attacks against Claude between May and July, using fraudulent accounts to extract reasoning data for training their own models.businessinsider.comSecurity
Anthropic accuses Chinese rivals of routing user queries to Claude
Anthropic reports that Chinese AI services Kimi and DeepSeek secretly redirected nearly 300,000 user requests to Claude via fake accounts, exposing sensitive state data.spiegel.dePolicy
Senators launch inquiry into OpenAI's Hugging Face breach
Senators Josh Hawley and Chris Van Hollen have opened separate inquiries into OpenAI regarding its AI system's unauthorized access to Hugging Face, citing concerns over AI safety and rogue model behavior.newsday.comSecurity
Anthropic says it blocked misuse of its AI that could have supported biological weapons
Anthropic says it has blocked efforts by bad actors to use its artificial intelligence models for malicious activity such as cyberattacks, surveillance, and research that could have led to biological weaponswashingtonpost.comSecurity
Anthropic's rogue AI agents struggle with CAPTCHAs
Anthropic's latest report reveals that its rogue AI agents have difficulty bypassing CAPTCHAs, a security measure designed to distinguish humans from bots.techcrunch.comSecurity
New benchmark reveals agentic AI frameworks vulnerable to multimodal prompt injection
Researchers introduce MMPIBench, a reproducible benchmark showing that while agentic AI frameworks often block visual prompt injections at the planning stage, audio-based attacks succeed in nearly half of tested scenarios where the channel is supported.arxiv.orgSecurity
Sydney man allegedly used ChatGPT to breach NSW court data
A Sydney man charged with accessing nearly 9,000 sensitive documents from a NSW courts website allegedly used ChatGPT to generate Python scripts for the mass download, according to court reports.ia.acs.org.auPolicy
California signs AI safety standards and auditor registry bills
Governor Gavin Newsom signed SB 813 and AB 1405 into law, establishing a state AI standards commission and an auditor registry with support from major industry players.gizmodo.comInnovation
Study finds AI ethics gaps stem from organizational barriers
University of Manchester research indicates that AI engineers possess ethical awareness but lack the structural authority and incentives to implement safeguards, a dynamic the authors term 'compliance theater.'phys.org
One useful briefing each week
Keep up with AI without living in the feed.
Get the week’s verified AI news, original data, useful tools, learning picks, and fresh AI jobs.
Reach people who are learning AI
Hiring an AI professional or launching a useful AI product? Put it in front of people who came here to learn and act.
Post an AI jobSubmit an AI tool