OpenAI will require members of its Trusted Access for Cyber program to secure their accounts with hardware-backed passkeys, tying access to the company’s most cyber-capable AI models to physical security keys rather than passwords or app-based codes.
From September 1, individual members of the program must enable OpenAI’s Advanced Account Security feature using a hardware-backed passkey to keep their elevated access; those who do not will revert to default model access. Trusted Access for Cyber gives vetted security researchers and organisations access to advanced AI capabilities for authorised defensive work, including vulnerability triage and validation, malware analysis, detection engineering, and patch validation. OpenAI is also tightening restrictions on high-risk entities and jurisdictions as part of the same effort to keep advanced cyber capabilities out of the wrong hands.
Hardware-backed passkeys store authentication credentials in a physical security key rather than syncing them through software or cloud services, so they cannot be copied or remotely extracted. Because the credential also verifies the legitimate service before authenticating, it resists phishing and adversary-in-the-middle attacks, in which a fraudulent site sits between a user and the real one to capture logins. The Advanced Account Security program lets users disable weaker fallback authentication methods that attackers might otherwise exploit.
Yubico is partnering with OpenAI on the rollout, offering existing OpenAI account holders a custom two-pack of YubiKeys at preferred pricing: a USB-C key with tap-based near-field communication and a low-profile key designed to stay plugged into a laptop. OpenAI already uses YubiKeys internally to protect employees and infrastructure. “We are introducing a new model for phishing-resistant security at scale for the AI ecosystem,” said Jerrod Chong, chief executive of Yubico.
The requirement makes identity a gatekeeper for AI capability itself, a pattern taking shape across the industry: Anthropic added identity verification for some Claude capabilities through Persona, and hardware keys have proven workable at enterprise scale, with T-Mobile deploying 200,000 YubiKeys across its workforce.
Beyond protecting individual researchers, the requirement changes the economics of account crime: credentials that cannot be phished, copied, or synchronised make compromised accounts considerably harder and more expensive for threat actors to create, validate, and resell at scale. For a program built around vetted access to sensitive capability, that raises the cost of the most obvious attack path.
Sources: iTWire
–
By Ji-seo Kim
