California Governor Gavin Newsom has officially announced a strategic agreement with Anthropic that fundamentally changes the state's approach to using artificial intelligence in the public sector. Under the terms of the partnership, all government agencies, as well as municipal and county authorities, will gain access to the AI assistant Claude at a 50% discount from market value.

Agreement Details and Educational Initiatives

In addition to the significant price preference, the agreement includes a comprehensive training program for government employees, technical support, and direct involvement of Anthropic engineers in optimizing workflows. This is the first time in U.S. history that an advanced language model becomes a standard tool for all levels of government administration through a single IT services portal, SITeS, of the California Department of Technology.

"Artificial intelligence is not intended to replace human labor in government. Its purpose is to increase employee efficiency, accelerate problem-solving, and deliver better outcomes for the people of California," Newsom emphasized in an official statement.

Practical Application and Cybersecurity

Several state government agencies are already actively using Claude for customer service, automating internal processes, and cybersecurity tasks. Particularly noteworthy is the model's application for scanning and fixing software code — a critically important function for ensuring the security of government information systems.

This agreement is a logical continuation of Newsom's consistent policy of integrating AI into government administration. As early as 2023, he directed state agencies to explore the potential of generative artificial intelligence, and in 2025, he signed an executive order on measures to improve government efficiency. In March 2026, the governor also updated AI procurement rules to account for civil rights requirements, privacy, and risk assessment for employees.

Context and Recent Events Surrounding Anthropic

It is worth noting that on June 9, Anthropic introduced two new models from the Claude family: Fable 5, positioned as a Mythos-class solution but safe for general use, and Claude Mythos 5 — a "private" base model with relaxed restrictions in certain areas. However, a few days later, developers were forced to restrict access to these tools for foreign citizens following a directive from the U.S. government. Several experts link this to incidents of models being used by groups associated with China. At the end of the month, Anthropic also accused operators affiliated with Alibaba of distilling Claude.

Expert Commentary: California's decision is not just a pilot project but a systemic step that could serve as a model for other states and even federal agencies. However, the key challenge here is the balance between efficiency and safety: the widespread adoption of AI in the public sector will require unprecedented measures for data protection and algorithm control. Observing Anthropic's recent restrictions on foreign users, it can be assumed that national security issues will play an increasingly significant role in regulating such partnerships.