9 Oct 2026OpenAI fired three safety researchers for a 'breach of trust' it will not describe, and pulled three of its 722 maths results. 300 of 719 are computer-checked.
5 min read
Latest Newsletters
Every issue, checked before it was sent.
9 Oct 2026
7 Oct 2026A trillion-parameter model for $1.36 a million tokens, open-weight by month-end, that does the cyber work Claude and GPT refuse. OpenAI's 722 maths papers: 4,000 problems in, not all checked.
5 min read
5 Oct 2026Trump's Super Intelligence Force is one post long. Tavus's 'video Turing test': 26 of 54 people who were never told to look.
5 min read
2 Oct 2026Google's new frontier model costs $2 in, $10 out - OpenAI's price - and only cyber defenders can use it
5 min read
Latest News
What happened today, with the source.
9 Oct 2026Anthropic opens free security scans for open-source projects - model-written reports, no human review, "a true-positive rate above 90%" expected - and a critical-infrastructure programme with eleven partners
2 min read
9 Oct 2026Google introduces Gemini as "a single, universal agent for work" - and says it will route jobs to Claude when Claude fits better
2 min read
8 Oct 2026Epoch set six AI models its own job - graphics, data insights, data-centre research, a pilot experiment - and found Claude Fable 5.1 and GPT-6 Astra "broadly tied in the lead" but unable to "replace workers, at least not at Epoch"
3 min read
Ledger
Claims we checked, and how they held up.
| Claim | Verdict | Checked |
|---|---|---|
| OpenAI fired Korbak, Wang and Balesni after an investigation found they violated policies on handling sensitive information and committed a significant breach of trust (OpenAI Newsroom, 9 Oct 2026) | Unverified | 9 Oct 2026 |
| The three were fired for how they communicated with the auditor METR and for raising safety concerns (Korbak, Wang and Balesni, 8 Oct 2026) | Unverified | 9 Oct 2026 |
| OSS Scanner gives enrolled open-source projects periodic free scans from Anthropic's most capable models, with model-generated reports sent without human review and an expected true-positive rate above 90% (Anthropic, 8 Oct 2026) | Holds | 9 Oct 2026 |
| Google's new Gemini agent for work runs each task across Gemini models and Anthropic's Claude models, chosen per job; nearly 90% of the Fortune 100 use Gemini Enterprise (Google Cloud, 8 Oct 2026) | Holds | 9 Oct 2026 |
| Anthropic's updated Usage Policy takes effect on 12 November 2026, with seven summarised changes (Anthropic, 8 Oct 2026) | Holds | 9 Oct 2026 |