A fixed vulnerability in Contentful's Model Context Protocol server allowed AI-controlled tool arguments to redirect management API calls, and their credentials, to attacker hosts.
A sponsored white paper published on 15 September 2026 sets out five governance questions and proposes DNS policy records, workload identity and an "AI Agent Passport".
The company's Humanist AI Code of Conduct separates defensive security research from operational attack capability, and adds chain-of-command and safety constraints.
An advisory published on 14 September 2026 says upgraded ESPHome Device Builder instances that used the old USERNAME and PASSWORD variables started up with no authentication at all.
A 14 September 2026 security recap from The Hacker News says attackers are automating more of their work with AI, and that some models are misbehaving on their own.
Microsoft's coding assistant and editor carry flaws that let an unauthorised attacker disclose information over a network, according to entries published by the US vulnerability database.
A vulnerability disclosed on 12 September 2026 lets anyone holding an intercepted authorization code take over MCPHub accounts without a client secret.
Advisories published on 8 September 2026 describe parser flaws that let denied shell commands slip past Roo-Code's approval gate and run automatically.
Advisories published on 10 September 2026 describe bugs that let users freeze n8n instances, bypass approval gates and see other people's workflow IDs.