Banks wall off AI agents as credential brokers arrive — Oct 23

Big banks are revising their terms of service to prevent the use of autonomous AI agents in customer accounts, citing security concerns. Institutions such as Morgan Stanley, Bank of America, and Wells Fargo are taking steps to block agents like Meta's Muse, which has recently gained significant traction with over 10 million downloads in just 40 days. The banks argue that these agents pose risks by potentially accessing user credentials, which violates their guidelines, as many only permit read-only access.

In the realm of AI development, OpenAI is reportedly seeking to raise up to $30 billion at a valuation of $1.25 trillion, while DeepSeek is also in the process of raising approximately $15 billion despite facing a federal ban on its products. Meanwhile, Anthropic has made strides in AI safety by developing formal proofs for open-source libraries, revealing numerous bugs and vulnerabilities. The company has also donated its AgentSkills standard to the Agentic AI Foundation, aiming to establish trust and security guidelines for AI skills deployment.

Additionally, Microsoft has released findings from a simulation indicating that its MAIDXO orchestrator outperformed primary care doctors in diagnosing complex medical cases. The orchestrator achieved an 83% success rate compared to 36% for generalists, suggesting that AI tools could enhance medical decision-making, although further testing in real-world scenarios is planned.

Banks wall off outside AI agents