Anthropic said Tuesday it will make its Claude AI models available to the entire U.S. federal government — executive, legislative and judicial branches — for a symbolic fee of $1 per agency for one year. The package includes both its enterprise-grade offering and Claude for Government, which Anthropic and the General Services Administration say is cleared to handle FedRAMP High workloads for sensitive but unclassified work. The move follows OpenAI’s similar $1 deal announced last week and comes after the GSA recently added OpenAI, Google (Gemini) and Anthropic to an approved list of federal AI vendors.

What looks at first like a public-service promotion is also a textbook market-play: win early adoption inside government workflows, learn real-world use cases, and — crucially — build the kind of operational footprint that can lead to later contracts, integrations and, for vendors, long-term lock-in. That calculus is not theoretical. Anthropic and peers already hold lucrative defense relationships: this summer, the Department of Defense awarded Anthropic, Google, OpenAI and xAI prototype contracts with ceilings of up to $200 million apiece to accelerate military uses of frontier AI.

OpenAI’s $1 announcement last week focused on executive-branch agencies; Anthropic is pitching a slightly broader reach. According to Anthropic’s statement and a parallel GSA “OneGov” notice, agencies across all three branches can request access to Claude for Enterprise and Claude for Government today for the token fee, and Anthropic says it will provide technical onboarding support as agencies begin to deploy the tools. The government-grade Claude product is positioned to meet FedRAMP High and DoD-level authorizations, meaning it’s intended to sit on the more secure side of the unclassified spectrum.

Why does that matter? Federal procurement can be painfully slow. Getting a model into the hands of analysts, attorneys, staffers and civil servants — even on a promo price — can make the technology the default for drafting, summarizing, information retrieval and internal automation. Once workflows are built around a vendor’s APIs and tooling, switching costs rise. Analysts who study enterprise software have long seen free or nearly-free early offers as a deliberate strategy to seed adoption; several reporters and industry observers have made the same point about these AI giveaways.

Anthropic and the GSA both leaned hard on security as part of the pitch. Claude for Government’s FedRAMP High and DoD IL5/IL4 authorizations are highlighted as evidence the model can be used for “sensitive unclassified” tasks — things like intelligence fusion, regulatory review, or internal casework that demand stronger controls than a consumer product. Anthropic also pointed to existing partnerships that surface Claude in cloud marketplaces (for example, Claude in Amazon Bedrock) to underline its readiness for federal environments. For agencies wrestling with both usefulness and custody of sensitive data, that compliance story is the critical selling point.

The political backdrop: policy, bias worries, and the AI Action Plan

This is not happening in a policy vacuum. The White House and GSA have been explicit about wanting to speed AI adoption across government — the administration’s AI Action Plan and a flurry of executive actions in recent weeks stress American leadership and rapid adoption of commercial tools. At the same time, those same actions also carry ideological baggage: the administration has signaled that federal AI systems should be “free from top-down ideological bias” and explicitly criticized frameworks like DEI in some policy language, which has prompted debate among lawmakers, civil-liberties groups and scholars about how “bias” will be defined and enforced. That tension — between eagerness to adopt AI and demands that models conform to political expectations — is a central unresolved question as companies race to onboard federal customers.

Why companies are willing to give away access

A few practical reasons explain the $1 gambit:

• Footprint and lock-in. Government users are influential, and federal procurement can lead to large, long-term contracts and integrations with big cloud providers and IT vendors. Early adoption can seed later revenue.
• Data and product improvement. Working with real-world agency workflows helps companies improve models and build tailored features for heavily regulated industries.
• Political positioning. Being seen in the tent — approved on GSA lists and deployed across agencies — buys influence in policy debates over safety rules and procurement.

Points of friction and concern

That said, the rush raises several concerns that watchdogs, privacy advocates and some lawmakers have already flagged:

• Data governance and leakage. Even FedRAMP High environments require careful design to ensure sensitive information isn’t exposed in model logs or third-party pipelines. The devil is in the deployment details — not the price tag.
• Operational risk. Models can hallucinate, misinterpret legal texts or mishandle classified-adjacent material. Agencies will need safeguards, human review and clear rules of engagement.
• Political and ethical questions. When the administration insists models be “free from top-down ideological bias,” watchdogs worry about vague standards being used to exclude certain viewpoints or to demand model behavior that aligns with partisan preferences. Scholars and policy shops have warned this could politicize model audits and regulatory enforcement.

What to watch next

• Adoption vs. audit. Will agencies adopt Claude and ChatGPT widely, and if they do, how fast will independent audits and oversight follow? Expect senators, inspector generals and privacy offices to push for transparency — and for companies to push back on what they regard as proprietary model internals.
• Procurement outcomes. The $1 a year is a runway. The industry will be watching whether those pilots convert into paid contracts, deep cloud integrations, or multi-year defense prototypes. The DoD awards and GSA listings make that conversion plausible.
• Policy tests. If regulators or the White House require models to meet new ideological-neutrality tests, companies could face a binary choice: change models (and potentially degrade certain outputs), or lose access to lucrative public-sector markets. The way those tests are written will matter enormously.

Anthropic’s $1 Claude offer is both pragmatic and strategic: it’s a low-cost way to accelerate adoption inside the sprawling federal machine while signaling the company’s readiness — and compliance credentials — to handle sensitive work. But the move also deepens a broader debate about how the public sector acquires, vets and governs powerful AI systems at a time when political demands about model behavior are growing louder. For the companies, the reward is clear: influence, product learning, and a path to larger deals. For the public, the payoff depends on whether agencies pair faster access with stronger oversight.