AI news Nov 11: Anthropic IPO, AI agents and safety

Anthropic has started the roadshow for a proposed IPO, with shares expected to price at $70.00 to $78.00. Elsewhere, companies announced new AI agent tools, model deployments and safety measures spanning software, infrastructure and frontier AI regulation.
- Anthropic expects its IPO shares to price at $70.00 to $78.00.
- Google released A2UI 1.0 for interactive agent interfaces.
- Databricks expanded Genie into autonomous enterprise work.
- Google DeepMind reported formal verification of selected Signal code properties.
- Australia plans six mandatory frontier AI safety guardrails.
Anthropic starts IPO process
Anthropic has begun the roadshow for a proposed initial public offering of Class A common stock. The AI developer expects a price of $70.00 to $78.00 per share, has filed Amendment No. 2 to its Form S-1 registration statement and applied to list on Nasdaq under the symbol ANT.
Other company news included AWS opening its first cloud infrastructure region in Chile, backed by an investment of more than $4 billion, and Armada and Johnson Controls starting sovereign AI production in Dublin.
Agents move into products and enterprise workflows
Google released A2UI 1.0, the first stable version of its open protocol for generative user interfaces. It allows AI agents to respond with interactive screens such as forms, charts and cards, rather than relying only on text.
Databricks introduced an autonomous version of Genie that can handle long-running enterprise work, track progress, follow up on missing information and take action as new data arrives. The company also expanded Genie Agents across documents, tools and partners, and made Genie Ontology generally available.
ElevenLabs previewed new models, AI agents and deployment options at its New York Summit, while JetBrains brought its Air agent workspace to IDEs. Salesforce also released SFR-AutoR&D, a set of agents designed to run research loops end to end.
Models, code and AI infrastructure
Google DeepMind says AlphaProof Code formally verified selected security properties in Signal’s protocol code, generating more than six million lines of Lean proof in four weeks. The agents flagged about 1% of the codebase for a second human look, and the result does not remove the need for human review.
Moonshot AI’s Kimi K3.1 Pro Highspeed arrived on Amazon Bedrock with a 1-million-token context window. Google also launched the Gemini Decisions API for closed-set choices across text, images, audio, video and PDFs.
Additional releases included DeepL’s Aure voice model, TIER IV’s open-source autonomous driving models and Supermicro’s AMD EPYC 9006 servers.
Safety, research and regulation
Australia will introduce six mandatory safety guardrails for frontier AI companies and establish a regulator with powers to impose significant penalties. Google DeepMind also announced AlphaGenome 2, designed to predict how genetic variants affect specific human cell types and states.
In other research and industry developments, Google Research proposed a wearable-based CKM health coaching framework, while Quantinuum and RIKEN expanded their Japan quantum AI partnership. GitLab warned that Claude Code settings files can expose secrets, and AWS added web search for LangGraph dependency monitoring agents.

