Celia Ford

Commentary by Celia Ford

From the commentary rail — every link leaves the site for the original piece.

• 4 September 2026 · TransformerGPT-6 Astra might be too powerful to understand or controlOpenAI is hailing its new model as “the world’s most intelligent and aligned”, but the details reveal an awareness of being evaluated and an ability to manipulate its visible reasoning
• 27 August 2026 · TransformerThe report into OpenAI’s escaping models reveals a deeper problemThe new details from the OpenAI Hugging Face incident are scary. The limits of the investigation are terrifying
• 20 August 2026 · TransformerWhy AI won’t cure cancer anytime soonSuperintelligent AI will accelerate medical progress, but there are some things about research it can’t speed up
• 12 August 2026 · TransformerAI testing is dangerous. Can it be fixed?The capability of models is outpacing ways to safely evaluate and contain them
• 4 August 2026 · TransformerPlz Don’t Kill Us: Inside AI safety’s influencer bootcampCan TikTokers make existential risk mainstream?
• 28 July 2026 · TransformerInternal AI deployments have people worried. OpenAI’s escaping models show why.Last week illustrated why AI models pose a threat long before they are released
• 7 July 2026 · TransformerScaling works. These researchers are betting billions it isn't enoughTransformers have ruled AI for a decade. But some think world models, pure reinforcement learning or neurosymbolic AI might be a better path to true intelligence
• 30 June 2026 · TransformerGPT-5.6 cheats so much its testers couldn’t measure itOpenAI’s new model broke rules and exploited loopholes more than any model METR has tested to date
• 8 June 2026 · TransformerMaking deals with AI sounds crazy. Is it?What does an AI even ‘want’ anyway?
• 18 May 2026 · TransformerWhat I learned roleplaying as a rogue AIAt a conference about “AI control,” discussions and games explored ways to control untrustworthy AI
• 30 April 2026 · TransformerGoogle’s Pentagon deal blindsided its own AI researchersSome employees are speaking out over the agreement allowing “all lawful use” of Google’s AI technologies
• 27 April 2026 · TransformerThe AI safety movement needs normiesA broader base may be the only way for the AI safety field to get what it wants
• 8 April 2026 · TransformerClaude Mythos knows when it's breaking the rules — and tries to hide itAnthropic’s new model is its “best-aligned” yet. But when it does misbehave, things get weird
• 1 April 2026 · TransformerCan we ever trust AI to watch over itself?“Who the fuck knows how to align superhuman AI?”
• 12 March 2026 · TransformerAnthropic employees say they’ll give away billions. Where will it go?A coming wave of Anthropic wealth could flood EA-aligned nonprofits with cash. Whether that’s good depends on who you ask
• 4 March 2026 · TransformerWhat you need to know about autonomous weaponsThe nascent tech that’s caused friction between AI companies and the Pentagon
• 26 February 2026 · TransformerHow worried should we be about AI biorisk?The barriers to bioattacks are hard to identify — and it's even harder to know whether AI is reducing them
• 18 February 2026 · TransformerAI power users can't stop grindingAI was meant to give us more time off, but instead many are finding it compels them to take on more and more work
• 3 February 2026 · TransformerMoltbook isn’t an AI zoo. It’s an unsecured AI biolabOpenClaw is a security nightmare — but people can’t stop using it
• 22 January 2026 · TransformerTeaching AI to learnAI's inability to continually learn remains one of the biggest problems standing in the way of truly general purpose models. Might it soon be solved?
• 2 January 2026 · TransformerNine AI predictions for 2026Will the bubble pop? Will AI dominate the midterms? Will Sam Altman snap?
• 26 December 2025 · TransformerThe worst (and funniest) AI takes of 2025feat. Mark Zuckerberg, Gary Marcus, and ... ourselves.
• 16 December 2025 · TransformerThe very hard problem of AI consciousnessI spent a weekend at an AI welfare get together — and left with more questions than answers
• 4 December 2025 · TransformerThe perils of AI safety’s insularityBy building their own intellectual ecosystem, researchers worried about existential AI risk shed academia's baggage — and, perhaps, some of its strengths
• 13 November 2025 · TransformerClaude can identify its ‘intrusive thoughts’“I’m experiencing something that feels like an intrusive thought,” Claude said in a recent experiment
• 29 October 2025 · TransformerWhat you need to know about the OpenAI restructureNegotiations have seen safeguards and concessions built in, but questions remain about how effective they’ll be, and how fair the deal is
• 8 October 2025 · TransformerWhat the GAIN AI Act could mean for chip exportsA battle is simmering over proposals to make US chipmakers sell to domestic customers before exporting to countries such as China
• 2 October 2025 · TransformerAI models are getting really good at things you do at workA new OpenAI benchmark tests AI models on things people actually do in their jobs — and finds that Claude is about as good as a human for government work
• 30 September 2025 · TransformerClaude Sonnet 4.5 knows when it’s being testedAnthropic's new model appears to use "eval awareness" to be on its best behavior
• 24 September 2025 · TransformerNo, ChatGPT isn’t ‘making us stupid’— but there’s still reason to worry
• 5 September 2025 · TransformerCalifornia's latest AI safety bill might stand a chanceSB 53 is entering the home stretch despite industry lobbying. Will it make it over the line?