AI News Daily
AI developments selected from public sources
- 8
- major stories
- 13
- sources
- 4
- first-party stories
Lead story
Federal Reserve Financial Stability Report: AI becomes Wall Street's top financial risk
The Federal Reserve's Financial Stability Report shows that in a New York Fed survey, 87% of market participants listed AI as the top financial risk, up from 50% in the spring. The report said market enthusiasm for superintelligence is one of the factors pushing up stock valuations; as of October, debt raised for AI infrastructure had exceeded $680 billion, about 5 times the full-year 2023 total and roughly 31% of investment-grade corporate borrowing. Large tech companies and chipmakers also hold more than $3 trillion in forward leases and guarantees, GPU makers hedge the value of the chips they sell and invest in the same group of buyers, forming circular financing, and the largest U.S. banks' exposure to large tech companies, data center builders and their financing vehicles is close to $1 trillion.
Model releases / updates
Anthropic releases two models, Claude Fable 5.2 and Mythos 5.2
Anthropic released Claude Fable 5.2 and Mythos 5.2 on November 9, saying both have 30% faster output and 20% lower cost than their predecessors. On Artificial Analysis's Intelligence Index, Fable 5.2 scored 65, above Gemini 4 Carbon's 60 and Claude Opus 5.5's 59. Fable 5.2 is now available in the Claude app, Claude Code and major cloud platforms, at $8 per million input tokens and $40 per million output tokens; Mythos 5.2 is the same model as Fable 5.2 but has no restrictions in cybersecurity and biology, and is currently open only to enterprises that sign up for that program.
Industry news
Musk signs petition calling for Jensen Huang to return to Nvidia; over 14,000 signatures overwhelm servers
Musk publicly confirmed that he is signing a petition calling for Jensen Huang to return as Nvidia CEO; the petition website is under strain from a surge in traffic, and reports say signatures have exceeded 14,000, with the page showing about 12,000. Former Nvidia GPU design team member and current Tesla engineer Yun-Ta Tsai called on employees who want Huang to return to pressure the company by resigning, and said excellent GPU engineers are harder to replace than machine learning engineers. Investor Naval Ravikant questioned replacing founders, citing Apple firing Jobs and Uber pushing out Kalanick as examples, and Musk replied, "Well said."
Cursor announces findings of security incident investigation: Fireworks key leak led to data exfiltration and a malicious Claude Code extension being listed
Cursor announced the findings of its investigation into the October security incident: attackers used leaked Fireworks access keys to read coding agent trajectory data in a read-only storage bucket, fewer than 0.01% of customer API credentials were exposed, and they used this to gain access to some SaaS tools. The attackers also used the stolen credentials to plant a malicious copy of the Anthropic Claude Code extension in the Cursor VS Code extension marketplace cache; on October 29, up to 5,000 users who requested a specific cached version may have been affected, and the copy existed for about 11 hours. Cursor has moved all inference traffic off Fireworks, rotated all exposed credentials and rebuilt the extension marketplace, saying the incident is under control, with the last attack activity occurring on November 1 and having been blocked.
Anthropic CEO Amodei says there is huge uncertainty next year ahead of IPO
Anthropic CEO Dario Amodei said at a JPMorgan Chase investor event in New York that there is huge uncertainty about what will happen next year, and that compute is the company's deepest business dilemma. He said compute must be purchased one to two years in advance, and whether revenue will cover the cost by then is unknown, with risks in buying too much or too little. According to Bloomberg, Anthropic has committed nearly $901 billion, of which about $71 billion will come due before September 2027; the company's annualized revenue reached $81 billion at the end of October, up $10 billion from the end of September.
Anthropic engineering blog publishes 10 recommendations for safety evaluations
Anthropic says safety evaluations are critical for determining whether a model can be deployed, what safeguards need to be built, and where training and research should focus. Its engineering blog offers 10 recommendations for designing and running safety evaluations and reporting results.
Claude Code desktop version supports custom themes
Claude Code desktop can now build custom themes via mods, allowing users to change color schemes and add dynamic backgrounds; users can also simply tell Claude the look they want, or use the theme-examples plugin to get started quickly. The feature currently only supports local sessions in the desktop version, with support for cloud sessions coming soon.
Claude Code desktop version supports custom themes via mods
Claude Code desktop can now build custom themes via mods, modifying color schemes and adding dynamic backgrounds. Just tell Claude the look you want, or use the theme-examples plugin to get started quickly. The feature currently only supports local sessions in the desktop app, with cloud session support coming soon.