Anthropic has released Mythos 5.2 and Fable 5.2, AI models built for difficult technical tasks such as coding and research. Anthropic says they produce responses up to 30% faster than the previous versions; Fable is broadly available, while access to Mythos requires approval for certain uses.
Anthropic released new versions of its Fable and Mythos AI models, its first top-tier launches since CEO Dario Amodei called for slower development. The company says the models are faster and cheaper than earlier versions, while access to Mythos remains limited as Anthropic prepares for a possible stock market debut.
Meta is expanding Teen Account Protections worldwide, using AI to identify Instagram users who may be teens even if they entered an adult birth date. The protections limit who can contact them and what content they see, and Meta plans similar efforts on Facebook and Threads.
Meta released Muse Ember, a proprietary model designed to handle routine steps in AI agent workflows. In an independent evaluation, it scored nearly as well as Claude Haiku 5.5 while costing about one-third as much per test task.
Disney is partnering with Adobe to develop custom Firefly AI models based on its characters and visual styles. The multiyear deal also gives Disney early access to Adobe tools for its productions, parks, cruises and ESPN content.
Google says it has released four TPU systems this year, adding the 8tx for handling sensitive data and the 8i Max for work that needs more memory. Both use existing chips, and Google is pitching them to outside customers as it competes with Nvidia.
Apple now sells $50 add-ons that let Creator Studio subscribers keep using AI tools after their monthly allowance runs out. Logic Pro is the first supported app, and unused purchased access lasts up to a year.
Security researchers found 1,122 public Claude Code settings files that contained passwords or access keys. The credentials could be saved when users chose not to approve a command again, then exposed if the settings file was uploaded to GitHub.
Medium has launched TK, a free writing app that lets users ask an AI assistant for help while drafting. It can offer editorial feedback, find saved notes and help writers explain how they used AI in published work.
Nvidia’s board is investigating CEO Jensen Huang after he acknowledged an affair with an executive whom US intelligence identified as a Chinese intelligence asset. The investigation puts his leadership in question and has unsettled an AI industry that relies heavily on Nvidia’s chips.