AI Daily
TOOLIN DAILYNvidia Board Investigates Jensen Huang Over Access Controversy with Frontier Models
The Nvidia board launched an independent investigation into CEO Jensen Huang, while Anthropic's Claude Code faced permission boundary and litigation pressures, and OpenAI completed a new round of financing while teasing new products.
The essentials
Nvidia Board Launches Independent Investigation After CEO Targeted in Foreign Intelligence Operation
The Nvidia board announced an independent investigation into CEO Jensen Huang, triggered by his close ties with an executive allegedly recruited to spy on the company. The probe is overseen by independent directors and outside counsel, with Nvidia stating there is no reason to believe confidential information was compromised.
Source ↗Anthropic's Claude Code Faces Permission Leaks and Lawsuits
Goldman Sachs withdrew the tool after 1,122 public files leaked passwords due to configuration exposure. A US court ruled that Anthropic must face a lawsuit, with plaintiffs claiming that the Max 20x advertised usage does not match reality.
OpenAI Completes Financing, Releases Teaser, and Updates System Routing
Reports indicate OpenAI completed an 11 billion dollar financing round at a 1.25 trillion dollar post-money valuation. The official account posted a teaser for tomorrow, and some users were routed to the new Sol model.
In brief
Chollet on Scaling
Francois Chollet defined data and compute scaling as a generalization capability multiplier.
METR Report Readout
A report showed a 63 percent drop in progress per dollar spent, sparking discussions on the drivers of capability acceleration.
Compute and Power Contracts
Anthropic reportedly purchased 3.4 billion dollars in gas generators, and the SpaceX contract includes termination clauses.
Tencent and DeepMind
Tencent released the Terminal-RSI 27B agent pipeline, and DeepMind noted accelerated scientific discovery.
MuJoCo 4.0 Released
The physics engine released version 4.0, using speculative contacts to handle collisions of thin and high-speed objects.
Coding Benchmarks
Benchmarks showed performance gaps widen among certain models when instruction modifications are introduced.
City AI Ban Rumor
A flash update stated that Bloomington, Indiana, legislated a citywide ban on the use of AI.