Security practitioner slams frontier labs' "tabula rasa" approach to cybersecurity

thedealdirector · x · 2026-11-08

A security practitioner critiques OpenAI and Anthropic's "tabula rasa" mentality in cybersecurity:

Frontier labs only began investing heavily in security training data 6 months ago; once "Mythos" landed, the "AI breaks the game" moment arrived, yet labs ignored nearly every conventional industry approach.
Core argument: security is fundamentally about people, process, and systems—not just bug hunting. Most breaches come from insider risk and stolen credentials, not novel exploits; equating AppSec with the entire field is a dangerously one-dimensional narrative.
AI's real impact is democratizing zero-days, previously held by nation-state actors and elite ransomware crews, now available to anyone who can spin up an abliterated LLM.
Labs have been adversarial with key researchers; their strategy is arguably "create a new problem, become its solution, then take over the industry's tech layer." Somewhat deserved, as security vendors are the griftiest software category (non-functional products, integration-breaking tools, predatory pricing and lock-in)—but pretending everything after Mythos is a blank slate is irresponsible.
Conclusion: this strategy will fail.

More from Companies & People