Menu
SUBSCRIBE
SUBSCRIBE
Menu

Mythos

Teal light streaks racing down a digital tunnel, on BSA's plan to give defenders first use of AI cybersecurity models

BSA Proposes Giving Cyber Defenders Early Access to Powerful AI Models

, , , , , , , , ,

Qualified organizations would use the models to find and fix security flaws before those capabilities become broadly available.

 

The Business Software Alliance (BSA), an enterprise software trade group, proposed giving trusted cybersecurity organizations early access to powerful AI models before releasing them more widely.

Under the proposal, governments, AI developers, and cybersecurity organizations would form a voluntary partnership to oversee these phased releases. BSA’s announcement did not name a government body to run the partnership, any AI developer that had agreed to participate, or a timetable for creating it.

Defenders would get a head start

The proposal would apply to the most powerful AI cybersecurity models, which BSA said can find security flaws, determine which are most urgent, and exploit them with limited human direction. BSA cited GPT-6 Astra and Mythos as examples but did not say exactly how capable a model would need to be to qualify for the program.

BSA said the same capabilities can help defenders identify weaknesses more quickly or help attackers exploit them. Its proposal would give selected defenders time to locate vulnerabilities and coordinate fixes before attackers gain broad access to the models.

The early users would include cybersecurity providers, software companies, critical infrastructure operators, and organizations working to secure widely used open-source software. BSA also wants to deploy these capabilities more broadly through security platforms that organizations already use. The proposal says the partnership should eventually support both organizations protecting their own systems and providers protecting their customers, but it does not explain how that wider rollout would work.

BSA sets tests for early access

Organizations would need more than cybersecurity experience to qualify. BSA said applicants should show that they have followed applicable laws and accepted practices for reporting and fixing security flaws. The partnership would also examine whether they have the people and resources to use the models effectively and whether their work protects critical infrastructure or widely used technologies. BSA did not say what evidence applicants would need to provide. 

Applicants would also need to show that they could prevent unauthorized access to the models. They would have to disclose who owns or controls the organization and identify any other parties with substantial influence. The partnership would also consider whether the organization has ties to a country of concern that could make granting access risky.

Priority access would be temporary

The proposed partnership would first decide which AI models are powerful enough to require a phased release. Most AI security tools would remain outside the program because they do not materially expand what cybersecurity teams can already do.

The restricted-release period would end once similar models became widely available or the security risks no longer justified limiting access. BSA said participants should know the expected length of the early-access period, but its proposal does not set a maximum duration.

The full proposal also calls for clear application rules and explanations when an organization is denied access. It does not specify who would make those decisions or what steps governments and AI developers will take next to establish the partnership.

Anthropic CEO Dario Amodei standing before the White House, representing the company's cybersecurity collaboration with the U.S. government through Project Glasswing

U.S. Government Used Anthropic’s Mythos To Find Classified System Vulnerabilities Within Hours

, , , , , , , , , , , ,

The government participates in Project Glasswing, which gives vetted partners access to Mythos, the Anthropic model that has found more than 10,000 software vulnerabilities since April.

 

The U.S. government used Anthropic’s Mythos AI models to identify vulnerabilities in classified systems within hours, according to the Associated Press. It is the most sensitive disclosed application yet of Project Glasswing, Anthropic’s restricted program that gives vetted organizations access to its most powerful AI models to find software vulnerabilities before attackers do.

It is unclear whether the government used Mythos Preview, the original version launched in April, or Mythos 5, the latest version. The U.S. government suspended access to Mythos 5 on June 12 after citing a jailbreak identified in Fable 5, the consumer version of the same model, and halted access to both on security grounds.

Anthropic launched Project Glasswing, concurrently with Mythos, after determining that Mythos was too effective at finding vulnerabilities to release publicly. The company restricted access to a short list of vetted technology and cybersecurity companies, along with select government organizations. Partners found more than 10,000 high- and critical-severity software vulnerabilities in the program’s first weeks.

The government and Anthropic continue to work together despite the ongoing conflict between the parties. In March, the Department of Defense designated Anthropic a supply chain risk after Anthropic requested assurances its models would not be used for mass domestic surveillance or fully autonomous weapons. Anthropic challenged the designation in federal court and secured a temporary injunction. In June, the government separately ordered Anthropic to suspend access to Fable 5 and Mythos 5 due to concerns about access by foreign nationals. Anthropic complied while publicly disputing the evidence behind the order. The lawsuit remains active.

The White House held direct talks with Anthropic CEO Dario Amodei earlier this year about cybersecurity and shared AI priorities. Anthropic said it plans to expand Glasswing’s membership.