All Radar picks

X OPPORTUNITY RADAR / SHORT TAKE

The Vendor Runs the Border

Anthropic’s expanded cyber program makes powerful models available through three access tiers. Who gets to operate the checkpoint?

Information Overload
1 min readSources
AI-generated illustration of a faceless checkpoint operator above three lanes labeled Defense, Red Team and Specialized, with banks, aircraft and power lines behind them.
AI-generated conceptual editorial illustration; not documentary photography.

Anthropic has expanded its Cyber Verification Program, giving qualifying security professionals access to Claude Opus 5.5, Sonnet 5.5 and Mythos 5.1 with reduced blocking classifiers for authorized defensive work.[1]

Reuters reports that Glasswing partners found at least 129,000 verified software vulnerabilities between April and July. Anthropic’s own open-source scans found another 5,500 between April and October. More than 33,000 of the disclosed verified findings were rated critical or high. These are company-reported findings, not a count of presently vulnerable production systems.[2][3]

The three access levels are Defense, Red Team and Specialized. Anthropic says Specialized applicants receive an in-depth review with the U.S. government; existing Glasswing members transition without reapproval for current models. The usage policy still applies, and grants can be narrowed or withdrawn.[1]

Here’s the inference: the company is selling a tool and deciding who can operate its most capable version. Controlled access may be responsible distribution. It is also private governance, with government-assisted review for the Specialized tier.

The border here is a metaphor for vendor-controlled access. The permission slip is quite real.

CORRECTIONS & UPDATES.

Correction ·
Clarified that the 33,000-plus high/critical figure applies to the disclosed findings rather than only the April–July partner subset. Clarified the vendor-permission metaphor and that joint U.S.-government review applies to Specialized access.

FOLLOW THE SOURCES.

Checked Oct 6, 2026. Numbered references point to the evidence behind factual passages. Commentary is our own.

  1. Cyber Verification Program (opens in a new tab)AnthropicBack to cited passage
  2. Anthropic opens its most powerful AI models to more security teams (opens in a new tab)ReutersBack to cited passage
  3. Expanding the Cyber Verification Program (opens in a new tab)AnthropicBack to cited passage
AICybersecurityAnthropic
Back to the latest picks