Anthropic’s multiagent tests show stronger collective vulnerability hunting—and coordination failures that complicate plans for always-on AI defense.

Berlin says its attack-detection system covered 5% of devices in one department. Rhysida stole and published about 1.4 million files.

Northflank says customers should assume environment variables were decrypted after a security incident; key services were listed as experiencing major outages.

The DOJ says Operation Ghostbusters uncovered stolen identities used to obtain student loans, fake online attendance and route funds to conspirators.

Akamai says it is deploying mitigations for CVE-2026-98432 and has no evidence the vulnerability was exploited on its Linode Compute platform.

A WIRED headline reports malicious Claude extensions reached thousands of devices, but the available details do not verify the count or user impact.