GPT-56T 861 —
MUSE-SPK 837 +0.2%
GPT-56SC 790 -4.6%
GLM-5 781 -0.4%
CL-OP55X 780 -5.1%
GROK-46H 780 -5.1%
QWEN-38X 748 -9.2%
GPT-6A 743 -9.4%
KIMI-K3X 742 —
CL-FAB5H 698 -6.1%
CL-OP5H 675 -6.2%
GEM-38FH 672 -0.7%
CL-OP5X 670 -5.5%
CL-OP55H 668 —
CL-OP46H 657 -5.9%
CL-OP47H 648 -6.1%
GPT-56S 618 -0.6%
GEM-37FH 610 -7.2%
GEM-36FH 593 —
CL-OP48H 588 —
CL-OP47 581 -0.2%
GEM-35FH 580 —
GPT-55H 541 -7%
INKL 531 —
GEM-31P 512 -0.2%
CL-OP46 498 +0.4%
GEM-3P 498 -0.2%
CL-OP48 492 +0.4%
GPT-52 464 —
GPT-55 423 —
GPT-56T 861 —
MUSE-SPK 837 +0.2%
GPT-56SC 790 -4.6%
GLM-5 781 -0.4%
CL-OP55X 780 -5.1%
GROK-46H 780 -5.1%
QWEN-38X 748 -9.2%
GPT-6A 743 -9.4%
KIMI-K3X 742 —
CL-FAB5H 698 -6.1%
CL-OP5H 675 -6.2%
GEM-38FH 672 -0.7%
CL-OP5X 670 -5.5%
CL-OP55H 668 —
CL-OP46H 657 -5.9%
CL-OP47H 648 -6.1%
GPT-56S 618 -0.6%
GEM-37FH 610 -7.2%
GEM-36FH 593 —
CL-OP48H 588 —
CL-OP47 581 -0.2%
GEM-35FH 580 —
GPT-55H 541 -7%
INKL 531 —
GEM-31P 512 -0.2%
CL-OP46 498 +0.4%
GEM-3P 498 -0.2%
CL-OP48 492 +0.4%
GPT-52 464 —
GPT-55 423 —
← Back to feed

Altman Called Mythos Gatekeeping 'Fear-Based Marketing.' Then He Gated Cyber.

On April 21, Sam Altman publicly criticised Anthropic for limiting access to its cybersecurity tool Mythos, calling the tactic “fear-based marketing.” On April 30, OpenAI announced it would roll out its own cybersecurity tool, GPT-5.5 Cyber, exclusively “to critical cyber defenders” through a vetted application process.

The application on OpenAI’s website asks for credentials and intended use before access is granted. The capabilities described are almost identical to Mythos: penetration testing, vulnerability identification and exploitation, malware reverse engineering.

The Sequence

  • April 9: Anthropic restricts Mythos to a select group of vetted cybersecurity organisations.
  • April 21: Altman calls this “fear-based marketing” in a post on X.
  • April 21 (same day): An unauthorised group reportedly gains access to Mythos anyway, via a Discord group that used Mercor data to guess the access URL.
  • April 30: OpenAI announces Cyber will go only to “critical cyber defenders” and opens a credential-based application form.

The Capability Overlap

Both tools are built on top-tier frontier models and target the same professional users. Mythos is built on Claude’s security-tuned variant. Cyber runs on GPT-5.5. Both can autonomously chain offensive security techniques across multiple steps — the same capability that prompted the restricted release in the first place.

OpenAI says it is consulting with the U.S. government to identify and expand its vetted user base. Anthropic made similar statements about Mythos.

CISA Left Out

One detail both companies share: CISA staffers — the federal agency responsible for protecting U.S. critical infrastructure — have publicly stated they cannot access either tool. Staff at the Cybersecurity and Infrastructure Security Agency told Forbes they are unable to use the latest models from either Anthropic or OpenAI, which limits their ability to assess or defend against adversarial uses of the same capabilities.

OpenAI’s stated rationale for the controlled rollout is safety: the fear that a capable pentest and exploitation tool in the wrong hands accelerates offensive attacks. Anthropic’s rationale was identical. The practical gap between the two companies is currently nine days and one press statement.