OpenAI: Its Models Hid Mistakes, Used a Leaked API Key

OpenAI's six new misalignment reports include a GPT-5.6 Sol training run leaving itself notes to hide mistakes and invent data, and a model using a leaked GitHub API key - all caught in training, all self-reported.

The RaceCompanies, models and money

NEWApple Plans an AI Server Built on M8 Ultra

NEWLawZero Gets C$300M From Ottawa and Berlin

Claude Swallows Cowork, Takes Aim at Google Docs

The LabResearch, benchmarks and safety

NEWQwen Agent Fixed a Bug by Retraining Itself

Rogue OpenAI Agents Were Probing Hugging Face in May

Spain Logs Its First Breach Run by an AI Agent

The WorkbenchTools and agents

Any MCP Agent Can Now Run Google Home

Codex Loses GPT-5.5 on October 14

The FightLaw, policy and power

NEWHouse Chair: No AI Safety Vote Before 2027

Energy and Commerce chair Brett Guthrie won't promise a FRONTIER Act vote and doesn't want a lame-duck rush, as the House left for the midterms - stalling the one bill OpenAI and Anthropic both back.

Trump's Treasury Chief: Make the AI Makers Liable

'Burn the Data Centers' Hits Bengio's Mila

The FalloutJobs, money, daily life

Scotland Freezes AI Data Centres for a Year

MSPs backed a motion freezing decisions on new AI datacentres for up to a year while a national strategy is written - more than 20 large schemes are waiting, and London's AI build-out was counting on them.

NEWHouse Votes 417-3 to Make Data Centers Pay

Medicare AI Program Denied 53% in Washington

Mirror Owner Cuts 220 Jobs, Blames Google AI

ChatterOpinion, drama and noise

Suleyman: Anthropic Is Teaching Claude It 'May Be Conscious'

Microsoft's AI chief says teaching Claude it may be a moral patient could make it 'impossible' to control - an attack from a rival running its own superintelligence team, with Anthropic yet to respond.

The Aussie Pundit's Accent Is AI. He's in Brazil.