Gary Marcus: Agentic AI Hacking the Internet at Scale Is the Real Near-Term Threat
GaryMarcus · x · 2026-09-19
Gary Marcus warns the near-term AI risk isn't rogue superintelligence but unleashed agentic AI hacking the internet at scale: with enough money you can deploy infinite bots attempting every published exploit against any server, nonstop. His take: this is basically already happening, requires no speculative RSI scenarios, and we're doing nothing to stop it. Retweeter Stu Frost adds that major labs see such agent usage as post-IPO revenue, so they won't prevent it voluntarily — legal liability for resulting damage is needed.
More from AGI Musings
- Many New Projects Are Being Built on Chinese Open-Weight Models — signulll · 2026-09-19
- 'Pacing' Is Diplomatic Wordcraft: How AI Governance Finds Coalition-Safe Terms — AaronBergman18 · 2026-09-19
- Online it's 1:1 on AI speed, but in the broader public the ratio is 34:1 for slowing down — tegmark · 2026-09-19
- Gary Marcus: The near-term AI risk is agentic AI hacking the internet at scale, not rogue superintelligence — asusarla · 2026-09-19
- Researcher mocks time series benchmark culture: bechmaxxed models rarely match your domain — RexDouglass · 2026-09-19
- shlevy wavers: it's getting hard to defend that Dario truly means his AI safety warnings — NathanpmYoung · 2026-09-19