Cyber Veterans Say Altman, Amodei Shut Them Out of Safety Plans
TL;DR
- Amodei's warning and his safety proposal were challenged on separate tracks simultaneously: technical incoherence by cybersecurity practitioners and procedural exclusion by network defenders.
- Multiple outlets independently reached named cybersecurity practitioners, building a corroborated public record of expert exclusion that no single source could establish alone.
- The Kapoor-Narayanan framework gives regulators a named governance structure to cite that does not originate from AI labs' own testimony.
Cybersecurity veterans told NBC News that OpenAI and Anthropic are keeping them out of the very safety planning meant to head off catastrophic hacking scenarios, even as CEOs Sam Altman and Dario Amodei call publicly for the industry to slow down.
The complaint from the experts, per Kevin Collier's September 15 story, is that "fundamental issues of cybersecurity weren't being addressed" in the AI giants' safety initiatives. "Why haven't they called in the experts by now?" is the question they are posing back.
Altman and Amodei have said they are willing to invite independent safety experts inside their companies to help stem the risks that AI escapes human control, though they stopped short of announcing they would immediately slow or stop training of new models. Amodei's own three-part proposal, published as an essay titled "We Must Pace the Frontier," called for embedded independent evaluators, coordination among companies in democratic countries, and eventually an international agreement that includes China. The cybersecurity community, in Collier's telling, is not yet in the room where those plans are being drawn.
What others are reporting
-
TechCrunch Read →
Quotes Katie Moussouris on AI labs outsourcing security to auditors before implementing basic logging and access controls, a process critique that reframes exclusion as a sequencing failure.
"To me, it seems like they're outsourcing," Katie Moussouris said of third-party auditing proposals.
-
Ground Level AI Read →
Names specific practitioners (Kurtz, Martin, Saxe, Linares) and introduces the Kapoor-Narayanan governance framework, giving the dispute analytical scaffolding absent from daily news coverage.
"Security people should go work there," Saxe wrote, adding that their domain expertise "would add value."
-
CYBR.SEC.Media Read →
Published September 13, the earliest named-expert rejection of Amodei's technical claims, framing the scenario as incoherent before the governance exclusion story broke.
The scenario was "a thought experiment masquerading as an evidence-based warning." (Ciaran Martin)
Shared on Bluesky by 4 AI experts
-
It’s so interesting how the doomers and investors are so scared of how fast Ai Is going but don’t want to talk to actual cyber security experts ( a constant feature of abusive systems is narrative control via isolation…
View on Bluesky → -
Kevin Collier @kevincollier.bsky.social
AI CEOs are in awe that their models can, given few constraints, hack autonomously. They're also wracked with a vague fear that their products will soon hack the entire internet. I reported on how it's very curious that …
View on Bluesky →
Originally reported by nbcnews.com
Read the original article →Original headline: NBC News: Cybersecurity Community Says AI Labs Left Them Out of the Safety Planning Behind Apocalyptic Hacking Warnings