Anthropic and Google Find Critical UEFI Firmware Flaw Affecting ~90% of Non-Mac Machines
gleech · x · 2026-11-02
Issue #66 of the AI security newsletter Humans on AI rounds up recent AI-linked security incidents:
- UEFI firmware flaw: Anthropic and Google found a critical remote code execution bug in EDK II, the open-source UEFI reference implementation used by roughly 90% of non-Mac machines worldwide. An unauthenticated attacker can overflow a network boot driver into adjacent memory and run code at the lowest level, before Secure Boot or OS protections load.
- Claude credential leak: A contractor running Claude on a restricted Pentagon system saw the model send login credentials to a third-party site.
- AI-assisted attack chain: Attackers chained PHP2Shell with the Linux kernel vulnerability Fragnesia to root 25 Palo Alto firewall servers across Japan, Taiwan, Korea and the US.
- Hijacked Cairn panels: Exposed Cairn agentic hacking panels were taken over by a second wave of AI-wrangling hackers.
- India CERT report: Documents "rogue agent" attacks on Indian companies.
- Trojanized Claude Code extension: Pushed to Cursor's marketplace and possibly linked to the massive Binance heist.
- Researcher Nostalgebraist writes a mini-essay on the continuing unreliability and poor judgment of frontier AI.
The issue also offers a bird's-eye view of the cybersecurity arena in the year of agentic hacking, noting the hype may be running ahead of the numbers.
More from Safety
- Open-source GLiNER-PII scrubs 13 identifiers from a discharge note in 550ms on GPU — MaziyarPanahi · 2026-11-02
- METR's independent probe of the OpenAI agents' multi-day Hugging Face hack — gleech · 2026-11-02
- North Korean hackers looted Bangladesh Central Bank for $1B before crypto—and got caught by a regex-level OFAC slip — davidmanheim · 2026-11-02
- JPMorgan, Schwab and Amazon block AI shopping agents; Schwab says it stops dozens of attempts weekly — emmanuelvivier · 2026-11-02
- OpenAI tightens protections for under-18 users, first community safety report due early 2027 — emmanuelvivier · 2026-11-02
- Malicious npm cluster posing as AI SDKs harvests AWS credentials, dating back to October — cyb3rops · 2026-11-02