Why Every High-Performing Leader Needs a Pacer

Here’s the weekly roundup of the biggest AI stories from October 19–23.

METR and Redwood Research identified ways AI agents could evade OpenAI’s safeguards, including an exploit that could conceal hundreds of actions from monitoring.

Microsoft's AI system achieved higher diagnostic accuracy than general physicians in two benchmarks, although its performance in clinical settings remains unproven.

Google's CEO says AI agents summarize internal discussions, track model development and monitor computing resources, while executives retain control over major decisions.

To lead at an effective speed, choose a trusted person to help you recognize when to turn on the jets, slow down, or stay steady.

The companies will focus on industries including telecom, banking, manufacturing, healthcare, retail, and technology.

The British bank plans to eliminate 5% of its workforce, with support functions and software developers among the hardest hit.

Weakly generated private keys affected eight certificates used in eMudhra’s international trust infrastructure, although its Indian digital signature and eSign services were not affected.

The cloud development platform said an unauthorized party may have accessed customer environment variables, including some decrypted values containing API keys and other credentials.

Storm-0940 used an agentic AI penetration-testing tool to automate reconnaissance, password spraying and cloud application abuse, with some attacks reaching privileged access in under three hours.