Latest in
Artificial Intelligence
Wednesday, October 21, 2026
Satya Nadella calls for a built-in pause button for AI systems
The growing crop of AI assistants you can run from a text thread
Why Anthropic is taking its AI agents off the live internet during internal testing
TypeSafe’s fast-rising Jev lands $870 million after a viral debut
More from Artificial Intelligence
Anthropic widens startup access with free Claude Team seats and API credits
Anthropic is expanding its Claude for Startups program with a free year of Claude Team, $1,000 in API credits, and access to tools meant to help young companies build on Claude.
LibreOffice treats the absence of AI as a privacy feature
The open source office suite says it will keep generative AI out of its default installation for now, arguing that local-only software gives users stronger control over sensitive documents.
Pinterest adds AI beauty guides that turn inspiration into salon-ready plans
Pinterest is adding a feature that translates beauty Pins into practical guidance, including the terms to use with stylists, estimated time, price range, and upkeep needs.
Mistral’s latest model is a bet on European AI without waiting for open weights
Mistral has introduced Mistral Large 4, a 1-trillion-parameter multimodal model that is available now through a guarded endpoint and slated to become open-weight after safety testing. The release shows how the company is trying to compete with both closed and open rivals while staying relevant to enterprise buyers.
OpenAI’s $7 billion share buyback gives employees cash while IPO timing stays unclear
OpenAI has reportedly bought back $7 billion in employee shares at an $852 billion valuation. The move provides liquidity for staff, but it also suggests the company may not be ready to go public soon.
Anthropic will add watermarks to Claude’s text to meet EU transparency rules
Anthropic says new Claude models will automatically watermark generated text and files so they can be identified by other systems, with older models to follow.
Why vulnerability response now depends on inventories, not just faster scanning
AI is speeding up both attack research and defensive patching, but the biggest bottleneck often comes after disclosure: finding every affected image, dependency and workload before remediation can begin.
ChatGPT now taps Apple Health and medical records across regular chats
OpenAI is broadening ChatGPT’s Health feature so connected medical records and Apple Health data can shape responses in any conversation, not just within a dedicated health area.
Minnesota’s anti-deepfake law can take effect while xAI challenge continues
A federal judge refused to pause Minnesota’s new ban on apps that create nonconsensual sexualized images, allowing the law to take effect even as xAI continues its lawsuit.