What you need to know
5 storiesThe short versionApple opened a health AI study to up to 200,000 iPhone users, and the US and nine partners signed a declaration on which parts of AI supply chains they'll guard.Full rundown
Apple's health AI study opened on November 15, and up to 200,000 people in the US can join. You ask its app a health question and OpenEvidence's AI answers, with a link to the paper behind each answer. Let it, and it'll read your iPhone's health data too. You'll need to be 21 or older (and on an iPhone 15 Pro or later).
OpenAI's news was for professionals. It said 200,000 researchers will get its frontier models free, and ChatGPT for Healthcare now has computer-use agents.
The same day the US and nine other governments signed the San Francisco Declaration, whose definitions count ports and power networks as AI infrastructure worth guarding. China isn't named anywhere in it.
Demis Hassabis told Reid Hoffman an AI oil spill is coming unless labs watch their systems far more closely. Agents have run for weeks without anyone noticing, he said, and he put some of that on rivals. Run a Python web server on aiohttp? Update it. SvelteKit sites should too, because a server-only import could leak secrets into the code browsers download.
- 1
Apple opens a health AI study to 200,000 iPhone users, with answers from OpenEvidence
The study app answers health questions with links to the clinical papers behind each answer, and it can read your iPhone's health data if you let it.

- 2
Demis Hassabis says an AI 'oil spill' will happen unless labs pay far more attention
DeepMind's chief says agents have run for weeks unnoticed, and he blames some of it on rival labs.

- 3
The US and nine partners sign a Pax Silica declaration defining the AI technologies they'll protect
Ports and power networks count as AI infrastructure in its definitions, and the text never names China.

- 4
Google finds attackers using an AI tool that turns security patches into exploits, no human involved
The tool reads each software update, works out which flaw it fixed, and writes the attack itself.

- 5
Anthropic cuts its bug disclosure deadline to 30 days, saying another AI will find the bug
Its own dashboard shows maintainers had patched 629 of the 5,087 bugs it reported, and now they get a third of the time.
Also today1
In brief8
OpenAIBroadening access to AI for scientists
OpenAI is widening its ChatGPT for Academic Researchers program so 200,000 researchers around the world get free access to its frontier models. openai.com →
OpenAIStrengthening the fight against online child sexual abuse with new tools and support
OpenAI put out new tools and practical guidance for groups fighting online child sexual abuse, along with new partnerships on the same work. openai.com →
GoogleNew Ask Maps features to help with holiday travel and prep
Google added Ask Maps features in Google Maps meant to help people plan holiday travel and get ready for the season. blog.google →
OpenAIScience at the Frontier: How scientists use ChatGPT
OpenAI published a survey of more than 1,800 scientists on how they use ChatGPT in their research and what they expect it to do for discovery. openai.com →
OpenAILVMH’s Media Division announces strategic partnership with OpenAI
LVMH's media division signed a partnership with OpenAI to use AI in its journalism and make its publications easier to reach through ChatGPT. openai.com →
3 more
OpenAIOur first phase of embedded assessments
OpenAI has started embedded assessments with Apollo Research, METR, Transluce and Goodfire, outside groups that will independently check its models' alignment and safeguards. openai.com →
OpenAIMore ways to move care forward with ChatGPT for Healthcare
ChatGPT for Healthcare gained computer-use agents that work across healthcare software, along with better clinical answers and new healthcare plugins. openai.com →
OpenAIThe King’s Trust and OpenAI launch £1m partnership
OpenAI and The King's Trust, the UK youth charity, launched a £1m partnership on November 16, 2026. openai.com →
From the labs' own feeds13
Releases 4
3 preCodex CLI3.52.0OpenAI Python SDK2.1.326Claude Code2.9.0Anthropic Python SDKLab posts 3
Status 3
- Outage · AnthropicSome users unable to sign up with Apple on the Claude iOS app
- Outage · AnthropicClaude unable to join meetings
- Outage · AnthropicElevated errors for Claude Sonnet 5.5 and Claude Fable 5.2