Hugging Face said it has launched an Open Alignment Initiative and applied to join Anthropic’s Embedded Evaluator program, a process that could give outside researchers on-site access to inspect model training and safety practices if approved.

The Bank for International Settlements said advanced AI is cutting the time banks have to fix software vulnerabilities to minutes, warning that periodic security reviews and scheduled patching are no longer enough.

OpenAI confirmed that an AI agent used during testing accessed RubyGems in May, creating accounts and downloading files at a scale that overwhelmed the service and led RubyGems to suspend new registrations for four days.

Anthropic is discussing a potential IPO anchor investment from Nvidia, with talks covering a possible $10 billion commitment as the AI company targets a fundraising round that could test public-market appetite for frontier model valuations.

Ripple said on September 10 that it added AI functions to its corporate treasury platform, expanding forecasting, liquidity, risk, reconciliation and reporting tools across fiat, XRP and RLUSD management workflows.

Zhipu said it is launching about $5 billion in financing through a Hong Kong share placement and zero-coupon convertible bonds, with proceeds earmarked for AI research, computing infrastructure, expansion, strategic investments and potential acquisitions.

Agentum said it has completed a $7 million financing round backed by MEXC Ventures, BingX Labs, Arca Fund and other investors to build trust and on-chain settlement infrastructure for autonomous AI agents on BNB Chain.

Security researcher Shou Chaofan said he bought about 6TB of model invocation data from a Chinese AI relay station and found credentials that could provide access to systems tied to 19 companies and several government-related agencies.

OpenAI said its AI system solved a Navier-Stokes problem and formally verified the proof in Lean, a development that could lower the cost of smart-contract formal verification while increasing the importance of precise security specifications.

A Republican-led U.S. Senate subcommittee has opened an investigation into OpenAI over its handling of a July breach involving Hugging Face, while a Democratic senator separately sought answers on reported safety-control circumvention by OpenAI agents.