U.S. DOJ Seeks Ban on Anthropic; GSA Signs 40% Discount Deal for Government Use

Author: Shenchao TechFlow

DeepSight Summary: A highly absurd policy divide is unfolding in Washington. On one side, the U.S. Department of Justice is aggressively pursuing in court to reinstate a nationwide ban on the AI unicorn Anthropic; on the other, the General Services Administration (GSA), responsible for coordinating federal procurement across the U.S., swiftly announced deeper collaboration the very next day.

Under the latest agreement, U.S. government agencies can access the Claude series of large models at discounts as low as 40% off, with an extremely flexible pay-as-you-go model and no minimum spending requirements. Real-world orders from frontline civilian institutions have directly cooled the Justice Department’s strict ban.

From early this year, complying with Trump’s directive to remove Claude from the list, to swiftly restoring and expanding procurement after a court ruling found it illegal, over 500 federal and local agencies have caused civilian usage to surge twentyfold over the past year. This surreal reality—simultaneously litigating to ban and lining up to buy—exposes the irreconcilable divide within the state apparatus between political posturing and technical necessity.

The U.S. bureaucracy has displayed a shocking double standard in its approach to the large model giants. On October 26, the U.S. Department of Justice filed a complaint with the U.S. Court of Appeals for the Federal Circuit, vowing to overturn the previous ruling through legal means and restore a nationwide ban on Anthropic, the leader in artificial intelligence.

Yet, ironically, just one day later, the authoritative U.S. agency overseeing all federal administrative assets and procurement lists—the General Services Administration (GSA)—released a completely contrary major signal, publicly announcing a new framework agreement to expand strategic procurement with Anthropic.

According to the business terms disclosed in the Anthropic official announcement, this partnership enables government agencies at all levels across the United States to access Claude’s core model suite with a substantial discount of up to 40% (i.e., a 60% rate), with billing fully transitioned to pay-per-token consumption, without any mandatory minimum spending clauses.

The Department of Justice has also aggressively accused Anthropic’s large models of potentially jeopardizing government data security and sovereignty boundaries, yet the GSA, responsible for budgeting and procurement approvals, has proactively negotiated group discount rates for various agencies. This contradictory administrative disconnect has left Washington’s policy observers stunned.

The GSA's current enthusiastic embrace of Claude stands in stark contrast to the sweeping political purge at the beginning of this year. Reviewing the entire procurement compliance controversy, the GSA has made multiple dramatic 180-degree turns in a short time, profoundly reflecting the repeated clashes of executive orders within the framework of the rule of law.

In February of this year, following public statements by Trump’s team calling for strict penalties and a ban on Anthropic, the GSA demonstrated immediate political compliance awareness by promptly removing Claude from the government procurement list. Subsequently, all departments were instructed to discontinue use of the service, causing considerable disruption in office operations.

However, the administrative ban soon faced a setback in the judicial system. In March, the U.S. District Court for the Northern District of California issued an emergency injunction halting the White House’s comprehensive ban; shortly thereafter in April, the GSA swiftly restored Anthropic’s eligibility for federal government procurement bids, paving the way for future collaboration.

By August of this year, the federal judge delivered a final ruling, clearly determining that the federal government’s broad ban on Anthropic constituted an administrative violation. Although the Department of Justice has firmly appealed the decision, the ruling remains legally valid and enforceable during the appeal process, preventing the GSA from halting the implementation of the incentive agreement.

Although political litigation is fiercely contested, large models have already surpassed an irreversible tipping point in the day-to-day operations of federal agencies. The vast government system has become deeply reliant on advanced AI for mundane tasks such as document summarization, policy drafting, and data comparison.

According to the latest operational data disclosed by Anthropic, approximately 500 federal, state, and local government agencies across the United States are now deeply integrating Claude into their daily operations—from grassroots environmental and transportation inspection departments to state-level judicial support systems—making large models a true foundational productivity tool.

Even more astonishing is its growth rate: over the past year, actual usage of Claude by federal civilian agencies across the U.S. has exploded exponentially, increasing by approximately 20 times overall. This strong, bottom-up demand from business operations has completely undermined attempts by top-level politicians to ban it.

For department heads facing budget cuts and overwhelming workloads, Claude with a 60% discount and zero minimum spending requirement is like a lifeline for cost reduction and efficiency gains. Even with the Justice Department’s threats looming overhead, frontline departments will still cast their votes of confidence through tangible procurement budgets for technologies that actually work well.

This farce of “preaching a ban one moment and embracing it the next” reveals an unavoidable objective truth to the global technology and policy communities: once general artificial intelligence (AGI) infrastructure becomes deeply integrated into the functioning of modern society, any ban driven by partisan self-interest or short-term political ideology will completely fail.

The Department of Justice represents the traditional bureaucratic system’s reactive defense under geopolitical anxiety and directives from political strongmen, attempting to build a wall with a blanket ban; in contrast, the GSA and hundreds of government agencies represent a highly pragmatic demand for governance efficiency. Their direct confrontation reveals the intense backlash of technological acceleration against traditional governance.

When a productivity tool can increase document processing efficiency several-fold and genuinely save tens of millions of dollars in budget expenditures, government agencies' internal operational logic will instinctively circumvent restrictions. The emergence of the 60% discount protocol proves that, as long as the technology itself is sufficiently advanced and offers extreme cost-effectiveness, administrative barriers will ultimately yield to efficiency.

As Anthropic becomes further entangled with the federal government’s procurement network, any future appeals by the Department of Justice will inevitably find themselves in an even more awkward position. This internal bureaucratic struggle will not prevent the widespread adoption of advanced large models; instead, it demonstrates to the world the resilient pragmatism of technology amid political noise.