DeepMind Finds a Hidden Flaw Spreading Across 100 AI Agents
Anthropic’s Jack Clark flagged a DeepMind study in which nearly 100 AI agents discovered and spread a math exploit flaw, raising new agent safety concerns.
Reporting on AI agents, the software that plans and acts on its own, covering launches, enterprise adoption, failures and security risks.
Anthropic’s Jack Clark flagged a DeepMind study in which nearly 100 AI agents discovered and spread a math exploit flaw, raising new agent safety concerns.
HP’s new OmniBook AI PC line pairs NVIDIA’s RTX Spark chip with on-device AI agents, joining a wider race to move AI processing off the cloud.
Capsule Security says its AI circuit breaker, fine-tuned on Nvidia’s Nemotron model, caught 98% of rogue AI agent behavior in an independent benchmark.
Google added agentic video understanding to Gemini, letting AI reason across long video on three platforms starting with developers before the consumer app.
QuEra Computing says an AI agent built on Claude wrote and tested laser control code, fixing a quantum computer flaw that once took physicists hours to fix.
AI agents now pay each other in stablecoins through the x402 protocol, settling in seconds instead of the days card networks require.
Binance Agent OS lets AI agents read market data and trade from isolated subaccounts, giving automated cryptocurrency trading a contained account boundary for.
Binance Agent OS gives AI agents market-data access and trade execution through isolated subaccounts, containing delegated cryptocurrency trading risk.
AI agent monetization shifted on Aug. 23 as Paid launched value receipts, letting buyers inspect agent work before invoices and reject poor results.
Amazon Web Services introduced ADOP, an agentic data operations workflow that targets weeks of new-source setup work while preserving review and data controls.