Marianne Sison

Visa’s live pilot lets AI agents complete approved supplier payments as banks restrict third-party account access. Greater financial autonomy depends on enforceable permissions and clear procedures for disputed losses.

METR and Redwood Research identified vulnerabilities that could conceal AI agent activity from OpenAI’s safety monitors. OpenAI says it has patched the affected safeguards, but researchers have not independently verified the fixes.

AI agents reached bank accounts while a China-linked actor used agentic AI in a supply-chain breach. Microsoft also reported a diagnostic system that outperformed general physicians on difficult medical cases.

Claude Duet checks connected applications and prepares business tasks for approval. Its value depends on whether the time saved gathering information exceeds the effort users spend verifying its recommendations.

AI-assisted research helped turn longstanding PHP weaknesses into working exploits. The emergency response exposed how limited maintenance resources and compatibility concerns can delay repairs as vulnerability research accelerates.

AI-discovered PHP flaws triggered emergency patches and website shutdowns worldwide. Google DeepMind employees resigned over military AI work, while Anthropic’s CEO defended the company’s safety record.

OpenAI bought Orchestra Bio to push AI deeper into drug development; Anthropic proposed a federal safety bill; Claude-designed toxin binders moved into lab testing; frontier AI compute costs surged.

A neuron that responds to cats and cars may recognize a shared visual pattern. Goodfire’s research challenges assumptions about AI interpretability, but explaining entire systems remains an open problem.

Anthropic’s proposed federal AI safety bill would give Commerce a route to seek court-ordered model suspensions. Its catastrophic-risk thresholds raise questions about earlier intervention and the independence of developer-funded evaluators.

DHS requires federal civilian agencies to remove covered DeepSeek products, including third-party implementations. Software vendors may face closer scrutiny of the models inside their products and how customers access them.

SemiAnalysis forecasts roughly 50 GW of additional global data center capacity in 2027. Active construction and long-term leases support the outlook, while electricity supply and chip availability could limit how much becomes operational.

A hacker used AI agents to breach more than 100 company sites, Microsoft introduced controls for risky enterprise agents, and OpenAI accelerated fundraising ahead of Anthropic’s expected IPO.