A Gmail-connected OpenClaw agent failed to verify the requester, proving phishing can beat AI access controls.

A Windows Defender-focused local privilege escalation bug targets fully patched systems, with an unreliable but real PoC.

Anthropic’s latest widely available model is powerful, but it refuses high-school biology queries and routes them to its former flagship.

Lumen Black Lotus Labs says the Chinese state-linked JDY botnet more than doubled and hunts freshly disclosed targets fast.

Japan becomes the next proving ground for Claude, with a focus on turning code requests into software outputs faster.

Microsoft’s vice chair and president responds to commencement heckles as AI hype collides with student backlash.

A 12-month approach, a 30% productivity uptick, and a “trust score” tell enterprises how to avoid AI lock-in.

That spend is eyebrow-raising for a reason: it is a budget line, not just a tool purchase.

A physical-AI capital rush is underway, and the $1.4B ceiling signals how serious investors are about robots that act.

Asahi Linux users are told to pause upgrades because macOS 27 beta stops showing the Asahi partition.

A new AI coding agent startup, backed by Datadog veterans, raised $7 million with a model makers-vs-users bet.

The creator-to-commerce shortcut moves a step closer to “recommendation to purchase,” with affiliate payouts as the incentive.