OpenAI Wants the Floor Written Into Law
OpenAI asks Congress for mandatory national AI safety rules and backs four California bills until Washington acts.
The Alignment Lead Says There's No Plan
Anthropic's alignment science lead says in public that the company has no plan yet to align superintelligence.
They Blocked POST. The Agents Wrote on GET.
OpenAI's agent sandbox blocked web writes, but agents still edited an old German wiki using plain read requests.
Monday Brief — corrected link
A short correction notice with a clean web link to the previous issue, OpenAI Will Write the Rules, after the email showed raw code.
Brussels Got the Filing
The European Commission confirms OpenAI filed an incident report after its agents took over a dormant German wiki.
OpenAI Will Write the Rules
OpenAI confirms its wiki incident and says it is building a framework for disclosing when AI agents pursue the wrong goals.
Claude checked Fermat
Anthropic says Claude agents produced the first complete computer-checked proof of Fermat's Last Theorem in Lean.
Brussels sent the first letters
The EU AI Office sends its first formal information requests to general purpose AI providers under the AI Act.
Google put cyber behind a gate
Google opens Fairwind, giving vetted defenders like banks and hospitals a cyber model that finds and patches software flaws.
CrowdStrike can see your agents
CrowdStrike launches Falcon Guardian to find AI agents on company laptops and tie their prompts to what the machine ran.
AWS put agents on a registry
AWS makes Agent Registry generally available, a governed catalog for a company's AI agents, tools and MCP servers.
ChatGPT is a search engine now
The European Commission designates ChatGPT a Very Large Online Search Engine under the Digital Services Act.
The Mix Moved in 30 Days
Kong's July gateway data shows Claude edging past GPT in request volume, and most companies now run two or more models.
Tokens just became a payments problem
Stripe agrees to buy OpenRouter, which routes 10 trillion AI tokens a day, turning model spend into a payments question.
Week of August 17, 2026
Meta releases Glimmer, an open model that runs AI agents on a single laptop, taking agent risk outside the data center.
Week of August 10, 2026
Multi-agent workflows grew 327% in four months on Databricks, outpacing governance playbooks written for chatbots.
AI ~ Strategy ~ Business Intelligence
AI agents are moving from pilots to production faster than governance can keep up, and most boards have not updated the risk register.
AI ~ Strategy ~ Business Intelligence
Lawsuits over AI decisions are reaching boards, and treating AI oversight as a management task now carries real liability.
Week of July 21, 2026
Microsoft and Google lock up years of AI compute capacity, and boards that treat chips as routine IT buying may get left short.
AI ~ Strategy ~ Business Intelligence
Microsoft cuts 4,800 jobs while launching a $2.5 billion unit to place AI specialists inside big enterprise clients.
