OpenAI's most capable cybersecurity models now come with a bouncer, and the bouncer is the company itself.
In February 2026, OpenAI announced "Trusted Access for Cyber," an identity- and trust-based framework that places its strongest cyber-capable models behind verification while committing $10 million in API credits to accelerate cyber defense. [1] Its most cyber-capable reasoning model, the company said, can now work autonomously for hours on complex tasks — capabilities that strengthen defenders and, in the wrong hands, attackers. The gate is the company's answer to that double edge. In April, OpenAI said it was scaling the program, broadening access for vetted defenders. [2]
On X, this lands as a morality play. One camp calls it censorship — OpenAI hoarding power tools and deciding who may wield them. Another calls it safety theater — a verification step that inconveniences researchers while doing little against well-resourced attackers. Both readings miss what the document actually establishes.
What it establishes is governance. A private firm is now deciding, case by case, who is trusted enough to use offensively-capable software, on criteria it sets and can change. [1] That is a state-like power — the licensing of a dual-use technology — exercised without a statute, a regulator, or an appeal. The April post frames the expansion as empowering defenders, but expansion and gatekeeping are the same act seen from two sides: every door OpenAI opens to a vetted defender is a door it can close to someone it declines to vet. [2]
This is the divergence the paper keeps. X argues access as free expression. Mainstream coverage — Wired, the wires — reports the cyber-capability race between labs. The under-covered story is the quiet transfer of a public function to a private one: who may use the most dangerous category of model is now an OpenAI decision, documented on an OpenAI page. [1]
The gate may be wise. Concentrating that judgment in one company is still a choice with consequences, and it deserves more scrutiny than either the censorship thread or the safety-theater dunk provides. Until a feed reads the policy as governance, it is arguing about a doorway while a company quietly writes the rules for who walks through. [2]
-- ANNA WEBER, Berlin