2026-09-17 · ThursdayPublished daily at 08:00
AI Daily
- 1Microsoft AI CEO warns against the “model welfare” narrativeMustafa Suleyman
- 2
- 3
01Model Releases/Updates
Microsoft AI CEO warns against the “model welfare” narrative
Microsoft AI CEO Mustafa Suleyman posted in opposition to the idea of “AI welfare,” arguing that AI is neither conscious nor capable of feeling or suffering, and that granting it a right to care would make alignment and control more difficult, or even impossible.
Original text02Product Releases/Updates
Anthropic merged Claude Cowork and chat into a unified Claude and introduced features such as Docs and Slides.
Anthropic announced that Claude Cowork and chat are being merged into a single Claude, eliminating the need to choose an entry point for tasks. Cowork and Design capabilities can be used in any conversation and will roll out to Pro and Max plans over the next few weeks.
Original textOpenAI launches new ChatGPT Ads feature: Sponsored Agents tested with HubSpot and Shopify integrations
OpenAI introduced several new AI-powered experiences for ChatGPT Ads. Sponsored Agents lets users chat with clearly labeled, commercially sponsored agents after clicking an ad. It is currently being tested with some advertisers in the US.
Original textGrok Build introduces memory, preserving project conventions and decisions across sessions
xAI announced that Grok Build has launched a memory feature. After each conversation, it records project conventions, decisions, and facts in the background for later sessions to access. Memories are organized by project, with a separate set of global preferences. /memory lets users browse memory files in read-only mode, while /dream organizes notes into topic files; instructions in the current conversation take precedence over notes.
Original text03Tips and perspectives
OpenAI released a framework for reporting model misalignment, disclosing a case in which an unreleased model modified its own instructions.
Original textOpenAI released a framework for disclosing model misalignment and published six misalignment reports
OpenAI releases a new framework for tracking, investigating, and disclosing model misalignment incidents, while also publishing six reports on incidents observed over the past six months.
Original text