Key Takeaways
- Anthropic is launching a cyber‑defense initiative that provides up to $15 million in AI credits to state, local, tribal and territorial governments (SLTTs) to help them find and fix vulnerabilities before attackers can exploit them.
- California and Texas are among the first participants, with California already reporting the discovery of more than three dozen weaknesses using Claude models.
- Eligible entities can receive up to $100,000 in Claude credits (six‑month expiry) and receive technical training and scheduled support from Anthropic’s applied AI team.
- The program is separate from the restricted‑access Project Glasswing, which serves critical‑infrastructure partners.
- Officials stress that AI‑driven speed is essential to outpace threat actors who can now weaponize frontier models to uncover hidden flaws in minutes or hours.
Introduction and Participation
Cybersecurity officials in California and Texas announced on Thursday that their states are joining a new Anthropic cyber‑defense initiative aimed at helping state, local, tribal and territorial governments (SLTTs) use artificial intelligence to identify system weaknesses before attackers can exploit them. The effort follows months of collaboration with public‑sector cybersecurity leaders and marks the launch of an inaugural cohort, with additional groups expected to begin work over the summer. Both states emphasized the proactive nature of the partnership, noting that they intend to stay ahead of emerging threats rather than merely react to them.
Funding, Credits and Support Structure
Anthropic has pledged up to $15 million in credits to participating SLTTs, enabling them to access the company’s Claude Security, Claude Code and the latest Opus 4.8 models. Eligible entities—including special districts such as port authorities or utilities, as well as very large cities and counties—may apply for up to $100,000 in Claude credits that expire after six months. Beyond the financial incentive, Anthropic’s applied AI team will provide technical training and hold scheduled calls to assist selected public‑sector teams in effectively integrating the AI tools into their vulnerability‑management workflows.
California’s Early Results
California Chief Information Security Officer Vitaliy Panych, who leads the California Department of Technology, reported that his agency had already begun using Claude and uncovered more than three dozen vulnerabilities in a short period. In an emailed statement, Panych said, “We are not waiting for the threat to mature before we act… our teams are already using Claude across vulnerability management and security operations, and in early use one of our security engineers surfaced 40 vulnerabilities in a fraction of the time it would normally take.” The rapid detection underscores the potential of AI to accelerate traditional security processes.
Scale and Speed in California’s Defense
Panych highlighted the scale of California’s responsibility, noting that the state “defends more people and more critical systems than almost any government in the country.” He argued that the partnership allows California to meet that scale with the speed necessary to identify weaknesses before adversaries can exploit them, thereby setting a benchmark other states can follow. By getting ahead of threats, the state aims to maintain a proactive cybersecurity posture that can adapt to evolving attack techniques.
Texas Cyber Command’s Perspective
Texas Cyber Command, also part of the initial cohort, views the initiative as a way to sharpen its responsiveness and flank emerging threats. Timothy “TJ” White, the command’s cyber chief, stated in a prepared comment, “There remains a lot of uncertainty and unknowns about what fully mature AI capability will mean when it comes to cybersecurity, but we do know that any number of bad actors will try to operationalize this, bringing a degree of disruption which will be difficult to deal with at speed.” He added that the collaboration ensures Texas is not merely reacting to potential disruption but building the agility needed to outpace threats.
AI’s Rising Stakes in Cybersecurity
The announcement comes amid growing concern that frontier AI models from companies such as OpenAI, Google and Anthropic are lowering the barrier for attackers to discover hidden vulnerabilities. During a May congressional hearing, Samir Jain, Vice President of Policy at the Center for Democracy and Technology, warned that these advanced models can uncover flaws that have lain dormant for years, enabling nation‑state and criminal actors to identify weaknesses in government systems faster than agencies can patch them. This shift has heightened the urgency for defenders to adopt AI‑driven proactive measures.
Shrinking Exploitation Window
An industry report released in April reinforced the warning, noting that assumptions about cybersecurity timelines are now outdated. The window between vulnerability discovery and exploitation has collapsed to hours, and in some cases minutes, as AI accelerates both offensive and defensive capabilities. The report urged organizations to leverage AI for internal vulnerability hunting before malicious actors can weaponize the same technology, a principle that underpins Anthropic’s SLTT initiative.
Distinction from Project Glasswing and Eligibility
Anthropic clarified that the new SLTT program is separate from Project Glasswing, a restricted‑access effort designed specifically for critical‑infrastructure partners. While Glasswing remains limited to a select set of high‑impact sectors, the SLTT initiative opens participation to a broader range of government bodies, including special districts, large municipalities and counties. This inclusivity aims to spread AI‑enhanced cyber defense capabilities across the entire public‑sector landscape.
Application Process and Ongoing Assistance
Interested entities can apply for the Claude credits through a streamlined process, with Anthropic’s applied AI team providing hands‑on support, technical training and regular check‑ins to ensure effective implementation. The six‑month credit period encourages recipients to integrate the tools into ongoing vulnerability‑management cycles, remediate identified issues, and develop sustainable practices that extend beyond the initial award period. By coupling financial resources with expert guidance, Anthropic seeks to create a replicable model for AI‑assisted cyber defense.
Conclusion: Building Agility and Setting Standards
Both Panych and White emphasized that the partnership is not merely about acquiring new technology but about cultivating a mindset of agility and foresight. As cyber criminals accelerate their use of AI, governments must likewise accelerate their defensive capabilities. The California and Texas examples illustrate how early adoption of AI tools can yield tangible vulnerability discoveries, while the broader SLTT initiative aims to disseminate those successes nationwide. Ultimately, the program aspires to establish a new standard for proactive, AI‑driven cybersecurity that other jurisdictions can emulate and build upon.