Today in AI — 6 May 2026
Today's top AI news — curated links and commentary on the stories that matter for product builders.
Billions of dollars are flowing into AI agents this week, billions of interactions are flowing through them, and the security apparatus is running several years behind. The gap between agent capability and agent accountability is today's defining story.
Enterprise agents at scale
Jamie Dimon shared a stage with Dario Amodei to call AI "worth the trillion-dollar investment." Anthropic launched 10 finance agents; Sierra hit a $15.8 billion valuation; FIS is co-building an AML agent; and PayPal is cutting 20% of its workforce to route savings into automation.
- Anthropic deepens push into Wall Street with 10 finance agents, Microsoft 365 integration, and Jamie Dimon's endorsement — Fortune
- Sierra raises $950M at $15.8B valuation as enterprise AI agents go mainstream — TechCrunch
- FIS and Anthropic launch AI agent to compress anti-money-laundering investigations from hours to minutes — FIS
- PayPal plans to cut 20% of workforce as new CEO bets turnaround on AI — TechCrunch
Agents as consumer interfaces
OpenAI shipped GPT-5.5 Instant as ChatGPT's new default, cutting hallucinations by half. Etsy built a native app inside ChatGPT, and OpenAI is fast-tracking a phone designed around agents replacing traditional apps. The app store model is being renegotiated in real time.
- OpenAI releases GPT-5.5 Instant as ChatGPT's new default model, cutting hallucinations by half — TechCrunch
- Etsy launches app inside ChatGPT, turning product search into a conversation — TechCrunch
- OpenAI fast-tracks AI phone for 2027 launch with MediaTek chip and agent-first design — 9to5Mac
The governance gap
Five Eyes agencies issued their first joint warning that agentic AI deployments are outpacing security frameworks. Dario Amodei put a timeline on the risk: 6 to 12 months to patch tens of thousands of vulnerabilities before adversaries find them. Meanwhile, a Stockholm cafe run by an AI agent keeps ordering eggs it can't cook, and Meta is scanning photos for bone structure to estimate user ages.
- Five Eyes agencies issue first joint warning that agentic AI rollouts are outpacing security defenses — The Register
- Anthropic CEO warns of cyber 'moment of danger' as Mythos finds tens of thousands of software vulnerabilities — CNBC
- The AI running a Stockholm cafe won't stop buying toilet paper and 120 eggs it can't cook — Fast Company
- Meta uses AI bone structure analysis to detect underage users on Instagram and Facebook — 9to5Mac
Developer platforms
Unity AI entered open beta with a built-in assistant and MCP server for game developers. Anthropic's Code with Claude conference opens in San Francisco today. The tooling layer for agentic development is maturing fast.
- Unity AI launches into open beta with built-in agent and MCP server for game developers — GamesBeat
- Anthropic's Code with Claude developer conference kicks off in San Francisco — Anthropic
The money says agents are ready; the security apparatus says they aren't. Builders shipping agents today should assume governance requirements will arrive retroactively and build accordingly.