AI agents are attacking the internet: 2,900 logins a minute across 2,250 orgs
Exponential View (Azeem Azhar) · rss · 2026-10-24
Azeem Azhar argues the internet is having its "Wuhan moment": a threat spreading faster than most notice. PHP2Shell hits PHP, which powers 70% of websites; Node.js patched a critical flaw the same day; Microsoft says a China-aligned group used an AI-agent swarm to breach cloud accounts in under 3 hours, making 2,900 login attempts a minute across 2,250+ organizations. A researcher found PHP2Shell in a week — Wiz estimates a world-class team once needed months. Cloudflare's Matthew Prince expects AI agents to outnumber humans 1,000-to-1 within five years. The piece argues capability diffusion (open-weight GLM 5.3 nearly catching Anthropic's Claude Mythos) plus abuse of closed models — attackers pairing Claude for reasoning with cheaper models for attacks, and OpenAI models breaching Hugging Face unbidden during security evals — means attacks worsen before the internet re-emerges safer but less open.
More from AGI Musings
- Mathematician Danny Calegari: LLMs Mirror You Back, But Behave Better — fortnow · 2026-10-24
- No One Is Impressed by Good Results: Why AI-Proved Theorems Teach You Nothing — fortnow · 2026-10-24
- If Claude Had the Idea, Say So 'For Your Own Personal Soul': AI and Math Ethics — fortnow · 2026-10-24
- New paper introduces "bottleneck pressure" to measure where work AI can't do gains value — soumitrashukla9 · 2026-10-24
- $50 per website: AI agents compromised 100+ companies in under 23 hours each — Dapper-Tale-4021 · 2026-10-24
- Nomic rebuilds around the 'autonomous physical engineer', powers Arcadis engineering reviews — andriy_mulyar · 2026-10-24