Gary Marcus: Doom talk distracts from frontier labs' incompetent security engineering
GaryMarcus · x · 2026-09-11
Gary Marcus quotes veteran security engineer Niels Provos: doomsday discourse distracts from the fact that frontier AI companies, especially OpenAI, aren't doing security competently. Models don't control inference, networks, or credentials—those external control points matter for the loss-of-control argument—and labs lack incentives to stop their own agents from misbehaving.
More from Companies & People
- Anthropic employee gifts an AI influencer 6 months of the 20x Max plan — scaling01 · 2026-09-11
- BioNeMo looks like NVIDIA's 'CUDA for biology' that could lock in a decade of bio AI — AllThingsApx · 2026-09-11
- YC's New Batch: 200 Startups, a Quarter Hard Tech, Up 40x From the Low — garrytan · 2026-09-11
- Everyone ships more than ever, but revenue isn't keeping up, says Clairevo CEO — lennysan · 2026-09-11
- Anthropic Announces Fable 5.1 Build Days: Global Buildathons Sept 11–25 — claudeai · 2026-09-11
- Perplexity joins the Rust Foundation to improve how agents build with Rust — AravSrinivas · 2026-09-11