Gary Marcus dismantles Dario Amodei's claim that rogue agent swarms could take over the internet in six months
GaryMarcus · x · 2026-09-13
Gary Marcus, Nathan Hamiel, and Zack Korman push back on Dario Amodei's warning that rogue AI agent swarms could take over the internet within six months:
- Vague and implausible: "taking over the entire internet" is ill-defined; attackers would knock themselves offline, and motives/budgets don't add up.
- Technically hard: Cloudflare, Google, and AWS are centralized but heavily hardened; a UK AISI report says frontier models can currently only compromise small, weakly defended systems.
- Execution doubts: if each bot needs a frontier model, the attack must run in the cloud under provider monitoring; the authors quip that if labs can't prevent abuse of their own models, they should be shut down.
- Kernel of truth: many small sites remain poorly secured and will be attacked; as models optimize and local hardware improves, more mischief becomes feasible without giant data centers.
Verdict: the six-month timeline is "silly," but it's still a reason to restrict hard-to-monitor agents.
More from Safety
- Frog and Toad meme nails the HSM paradox: keys stay safe, but compromised servers can still sign — rickasaurus · 2026-10-31
- Former US Chief Data Scientist DJ Patil publishes memo on the AI cybersecurity reckoning — matt_slotnick · 2026-10-31
- Why an AI military takeover is unlikely: armies won't hand over the reins — S_OhEigeartaigh · 2026-10-31
- Heads Of Security: paper pinpoints and rewires attention heads behind prompt injections — jonasgeiping · 2026-10-31
- House Democrats resolution: no bailouts for failing AI companies — your data goes to bankruptcy court — alex_verem · 2026-10-31
- Amazon Echo kills local voice storage: everything goes to the cloud from March 28 — alex_verem · 2026-10-31