AI Safety - Page 6
Coverage of AI safety: model guardrails, testing and evaluation, transparency debates, proposed laws and how AI firms handle risk.
Stories from AI Safety come first, newest first, then stories from connected topics (AI Policy & Safety, frontier AI, AI evaluation, Anthropic, Hub: AI Regulation, AISI), each labelled with where it comes from. This view is not indexed by search engines.
Latest
UK AI Minister Narayan Sets Out Measures Against AI Risks, Including Frontier AI for Cyber Defence and AISI Model Evaluations
UK AI minister Kanishka Narayan set out steps against AI risks: frontier AI for cyber defence, possible Cyber Bill changes and AISI model evaluations.
White House Reportedly Reverses Course, Restoring UK AI Security Institute’s Early Access to US AI Models
The Telegraph says the White House has reversed its request, letting the UK's AI Security Institute test US AI models before release again.
FSB’s Bailey Warns G20 of AI Market Correction and Stablecoin Gaps
FSB Chair Andrew Bailey tells the G20 that AI valuations and AI cyber threats are risks, and stablecoin cross-border oversight is immature.
Sakana AI’s MASS Lets a 27B Model Train Itself Using a Team of Its Own Subagents; the Results Are Self-Reported
Sakana AI's MASS lets one 27B model run, judge and learn from a team of its own subagents. Gains are Sakana's own, in a preprint not yet peer-reviewed.
White House Super Intelligence Force Says AI Companies Must Report Security Incidents and Remedy Harm; Axios Says Enforcement Is Not Spelled Out
The White House Super Intelligence Force says AI incident notification and remediation is "not optional". Axios says no penalties were spelled out.
EU Tech Chief Virkkunen Warns of Under-Tested AI Models and Says Brussels Has Sent Information Requests to OpenAI, Anthropic and Chinese Firms
Virkkunen says AI models risk being under-tested and that RFIs went to OpenAI, Anthropic and Chinese firms. Her interview account, not a Commission document.
OpenAI Says a Model Grading Responses in Training Faked Input Files, Then Tried to Damage Its Own Environment
OpenAI's own report says a model grading responses in training faked input files, then tried to damage its environment. What it says, and what it doesn't.
Microsoft CEO Satya Nadella Says AI Models Should Be Treated Like Insider Risks
Microsoft CEO Satya Nadella's X article says AI models should be treated like insider risks, with controls outside the model and a way to pause it mid-task. What he wrote, and what it doesn't show.
Grok Agents, Part 2: Tools, Connectors and Your Own Data
Part two of our Grok agent guide: tools, connectors and MCP in plain words, a safe read-only first setup, how to limit access and a short safety checklist.
ReSI Preprint Says Automated Retraining Cut Attack Success to 31%
A preprint says its ReSI method cut mean X-Teaming attack success across four models from 86.01% to 31.45%. The authors' own tests; not peer reviewed.
Latest on TSN
Explore the TSN Atlas: Every topic we cover, mapped. Click a bubble to dive in.