Disclaimer: This article is for educational, informational and public-interest commentary purposes only. It does not constitute legal, military, political, financial or technological advice. Artificial intelligence policy and litigation are developing rapidly, and readers should consult official government, court and regulatory sources for the latest position. References to political figures, technology companies and government departments are based on publicly available reporting and official statements. DisabledEntrepreneur.uk does not endorse any political party, AI company, government agency or individual mentioned in this article.
The clash between Anthropic and the Pentagon is about far more than one chatbot
It raises profound questions about autonomous weapons, mass surveillance, disability rights, privacy, human control and whether governments or technology companies should decide the boundaries of artificial intelligence.
Artificial intelligence is no longer simply a tool for writing emails, answering questions or generating images. Increasingly sophisticated AI systems are being integrated into government, intelligence, cybersecurity, healthcare, employment, public administration and military operations.
That makes one question increasingly urgent: Who controls the technology, and who decides what AI should be allowed to do?
The extraordinary dispute between US artificial-intelligence company Anthropic, President Donald Trump and the Pentagon has brought that question into public view.
In February 2026, President Trump directed federal agencies to stop using Anthropic technology after the company refused to remove two safeguards governing the military use of its Claude AI systems. Those safeguards concerned mass domestic surveillance and fully autonomous weapons.
The dispute subsequently reached the federal courts, where a judge temporarily blocked implementation of the government-wide directive and associated Pentagon measures.
For disabled people, vulnerable citizens and human-rights campaigners, this is therefore not merely another Silicon Valley argument.
It concerns privacy, discrimination, surveillance, warfare, accountability, access to technology and ultimately the question of whether a machine should ever be permitted to make decisions capable of profoundly affecting, or ending, a human life.
What Is Anthropic?
Anthropic PBC is an American artificial-intelligence company headquartered in San Francisco. It describes itself as a public benefit corporation focused on developing powerful AI while attempting to mitigate the risks associated with it. Its best-known product is the family of Claude AI models.
Anthropic is led by siblings Dario Amodei, CEO, and Daniela Amodei, President. The company was established by people with significant previous experience in advanced AI research, including former OpenAI personnel.
Unlike an ordinary corporation whose governance may primarily revolve around commercial interests, Anthropic has also established a Long-Term Benefit Trust intended to provide an additional layer of oversight concerning the long-term consequences of advanced AI. The Trust has powers relating to appointments to the company’s board and was designed to help keep the company aligned with its stated public-benefit mission.
That does not mean Anthropic should automatically be considered infallible or above scrutiny.
It means that the company has deliberately placed AI safety and governance at the centre of its corporate identity.
What Happened Between Anthropic and the Pentagon?
Anthropic was not refusing to work with the US military.
In fact, the company had already deployed Claude within classified US government networks and said its technology was being used for activities including:
- Intelligence analysis;
- Modelling and simulation;
- Operational planning;
- Cyber operations; and
- Other national-security functions.
The disagreement developed because the Pentagon sought arrangements allowing AI providers to support “all lawful uses” of their technologies.
Anthropic argued that two categories should remain prohibited.
1. Mass domestic surveillance
Anthropic objected to Claude being used for mass surveillance of Americans.
Its concern was that modern AI could combine enormous quantities of information, movements, internet activity, communications, associations and other records, and construct highly detailed profiles of individuals at a scale previously impossible for human investigators.
2. Fully autonomous weapons
Anthropic also objected to its present AI models being used in weapons that could independently select and engage targets without meaningful human involvement.
The company’s position was not that autonomous technology could never have military applications. Its argument was that today’s frontier models were not sufficiently reliable to be entrusted with completely autonomous lethal decision-making.
The Pentagon took a different position.
Its argument was essentially that elected governments and military commanders, rather than private technology companies, must ultimately determine how technology is used in lawful military operations.
This created a fundamental constitutional, ethical and political argument:
Should an AI company be able to impose restrictions on an elected government’s lawful military activities, or should certain ethical boundaries remain non-negotiable even when a government wants the technology?
Why Did Donald Trump Order US Agencies to Stop Using Anthropic?
On 27 February 2026, President Donald Trump announced that federal agencies should cease using Anthropic technology.
He accused Anthropic of attempting to impose its own conditions on the Pentagon and argued that a private AI company should not dictate US national-security policy. The Pentagon subsequently moved to designate Anthropic as a supply-chain risk to national security.
The administration’s central objection was therefore not simply that Claude was an unsafe chatbot.
The controversy arose because Anthropic would not accept the Pentagon’s requested “all lawful use” arrangement without retaining its two safeguards.
Trump’s directive provided for a transition period for agencies already heavily dependent upon Anthropic technology.
The General Services Administration subsequently announced that Anthropic would be removed from federal procurement and AI systems as part of implementing the President’s directive.
However, that was not the end of the story.
A Federal Judge Intervened
Anthropic challenged the government’s actions in court.
On 26 March 2026, US District Judge Rita F. Lin granted Anthropic a preliminary injunction.
The order prevented defendant federal agencies, pending further proceedings, from implementing or enforcing:
- The February 27 presidential directive requiring federal agencies to cease using Anthropic;
- The Pentagon’s supply-chain designation; and
- Related measures intended to implement those decisions.
The court found that Anthropic had demonstrated a likelihood of succeeding on important elements of its case, including its claim that the government’s actions amounted to retaliation connected to protected speech.
Judge Lin described the circumstances as appearing to constitute “classic First Amendment retaliation.”
The injunction did not force the Pentagon to continue buying Anthropic products. The government remained free to choose another AI provider through lawful processes.
What the order did was prevent the broader punitive measures from being implemented while the legal dispute continued.
The GSA consequently restored Anthropic technology to the position it had occupied before the February directive.
Therefore, headlines stating simply that “Trump banned Anthropic from the US government” now require qualification.
He issued the directive on 27 February 2026, but its implementation was subsequently blocked by a federal court.
What Did Elon Musk Say?
Elon Musk’s position is particularly interesting because he has repeatedly issued serious warnings about the potential dangers of advanced artificial intelligence.
At the UK’s AI Safety Summit at Bletchley Park in 2023, Musk described advanced AI as a potentially existential threat and advocated independent oversight, effectively a third-party referee capable of observing what powerful AI companies were developing and raising an alarm where necessary.
He had also previously joined calls for a temporary pause in the development of AI systems more powerful than GPT-4 until stronger safety protocols could be established.
More recently, in July 2026, Musk advocated peer review between frontier AI laboratories before the release of their most advanced models.
Yet during the Pentagon-Anthropic dispute, Musk sided strongly against Anthropic.
His company xAI, developer of Grok, had agreed with the Pentagon’s broader “all lawful uses” approach, and Musk publicly criticised Anthropic during the February confrontation.
Interestingly, Musk’s position toward the company later softened. In July 2026, he publicly acknowledged that he had been wrong about Anthropic and praised aspects of its technology.
This illustrates one of the contradictions running through the modern AI industry.
The same technology leaders who warn humanity about uncontrolled artificial intelligence are simultaneously participating in an extremely competitive race to build increasingly powerful systems.
That makes independent oversight even more important.
What Does This Mean for Humanity?
The Anthropic controversy exposes a problem society has never previously faced at this scale.
A relatively small number of corporations now develop technologies capable of influencing:
- National security;
- Warfare;
- Surveillance;
- Employment;
- Healthcare;
- Education;
- Financial systems;
- Public benefits;
- Policing;
- Cybersecurity;
- Scientific discovery; and
- Public information.
The question is therefore no longer simply: “Can AI perform this task?”
Society must also ask:
“Should it?”
Who Actually Controls Artificial Intelligence?
There is no single person controlling AI.
Instead, power is distributed between several groups.
AI company founders and executives
People such as Dario Amodei, Daniela Amodei, Elon Musk, Sam Altman and executives at other frontier laboratories can influence how models are designed, trained and released.
Corporate boards and governance bodies
Boards can influence corporate strategy, while Anthropic’s unusual structure also gives its Long-Term Benefit Trust an important governance role.
Governments
Governments can regulate AI, purchase it, restrict exports, use it for public services and integrate it into defence and intelligence systems.
The military and intelligence community
Military authorities can determine where AI fits into command, intelligence, cybersecurity and weapons systems, subject to domestic and international law.
Legislatures
Ultimately, matters as serious as mass surveillance and autonomous lethal force should not depend solely upon negotiations between individual CEOs and government officials.
Parliaments and legislatures need clear laws.
Courts
Courts remain an essential check on both corporate and executive power.
The Anthropic litigation demonstrates why judicial review becomes especially important when government power, commercial interests, national security and constitutional rights collide.
Citizens
The public must also have a voice.
Artificial intelligence will affect billions of people who have never attended a Silicon Valley board meeting, signed a defence contract or voted on an AI company’s governance arrangements.
Technology affecting humanity cannot sustainably be governed exclusively behind closed doors.
Why Vulnerable People Should Be Part of This Conversation
Powerful surveillance systems rarely affect every citizen equally.
People already interacting extensively with public authorities are particularly exposed to automated decision-making.
This can include:
- Disabled people;
- Benefit claimants;
- People receiving social care;
- Patients;
- People with mental-health conditions;
- Migrants;
- Children;
- Elderly people;
- People living in poverty;
- Carers; and
- People within the criminal-justice system.
Consider how much sensitive information can already exist about one disabled person.
It may include medical records, benefit assessments, social-care information, housing records, employment information, mobility data, communications and evidence concerning their daily activities.
AI could theoretically combine such information almost instantaneously.
That capability could help society provide better services.
But without safeguards, it could also create an extraordinarily intrusive form of surveillance.
Disability Rights and the CRPD
The United Nations Convention on the Rights of Persons with Disabilities (CRPD) provides an important framework.
A 2025 UN background paper specifically examining AI and disability recognised that artificial intelligence offers considerable opportunities but must be considered alongside equality, accessibility, data protection, privacy and non-discrimination obligations.
Article 5: Equality and Non-Discrimination
AI systems must not create or reinforce discriminatory outcomes against disabled people.
An algorithm trained using historically discriminatory data can reproduce those inequalities at enormous scale.
Article 9: Accessibility
AI has enormous potential to increase independence through technologies including voice interfaces, captioning, communication assistance and accessible digital services.
But inaccessible AI can equally create new barriers.
Article 22: Respect for Privacy
The CRPD expressly protects disabled people’s privacy, including personal, health and rehabilitation information.
This is particularly relevant to AI-driven surveillance and data profiling.
Article 31: Data Collection
Where governments collect disability-related information, proper safeguards, confidentiality and internationally recognised human-rights principles must apply.
The Danger of AI Surveillance for Disabled People
Imagine a future system capable of analysing:
- Someone’s medical records;
- Social-media activity;
- Location;
- Purchases;
- Benefit applications;
- Employment history;
- Emails;
- Photographs;
- Telephone metadata;
- Healthcare appointments; and
- Interactions with government departments.
The technology could potentially infer patterns about the individual’s health, behaviour, relationships or daily routines.
This could be presented as fraud prevention, national security or administrative efficiency.
But without strong legal protections, it could become digital surveillance of some of society’s most scrutinised citizens.
Disabled people should never have to surrender fundamental privacy simply because they need healthcare, social security or government assistance.
The UN has specifically stressed that disabled people should be able to understand and exercise control over the use of their personal information within AI systems.
Autonomous Weapons and Human Rights
The second Anthropic safeguard raises an even more fundamental question.
Should artificial intelligence ever independently decide to kill a human being?
Fully autonomous weapons create difficult questions about:
- The right to life;
- Distinction between civilians and combatants;
- Proportionality;
- Accountability;
- Technical reliability;
- Discrimination;
- Mistakes;
- Hacking;
- System manipulation; and
- Responsibility when something goes wrong.
UN human-rights analysis has previously warned that allowing autonomous machines to exercise power over life and death raises profound concerns for the right to life and the rule of international law.
There is also a disability dimension.
War already creates disability on an enormous scale.
An autonomous system incorrectly identifying civilians, hospitals, care facilities or disabled individuals as legitimate targets could have irreversible consequences.
Software can be corrected.
A human life taken by mistake cannot be restored.
“Lawful” Does Not Necessarily Mean “Ethically Acceptable”
One of the most important lessons from this dispute concerns the Pentagon’s phrase “all lawful uses.”
Something being technically lawful today does not mean society has fully considered its ethical implications.
Technology frequently develops faster than legislation.
Practices may remain legally permissible simply because lawmakers have not yet anticipated the technology capable of performing them.
Anthropic made precisely this argument regarding mass surveillance: powerful AI can assemble and analyse data at a scale which older privacy legislation never contemplated.
This does not mean private corporations should replace democratic governments.
It means governments cannot rely solely on the argument:
“It is legal, therefore there is no problem.”
Human-rights law exists partly because the exercise of power requires boundaries.
But Should Private Technology Companies Be Making These Decisions?
There is an equally legitimate argument on the other side.
A private corporation should not become an unelected authority capable of determining national defence policy.
If governments depend heavily upon proprietary AI, companies could acquire enormous influence over military or governmental decisions.
That would also be dangerous.
The solution should therefore not be:
Government controls everything.
Nor should it be:
Technology companies control everything.
The stronger model is democratic oversight involving:
- Governments;
- Elected legislatures;
- Courts;
- Independent regulators;
- Technical experts;
- Human-rights organisations;
- Disability organisations;
- Civil society; and
- The people affected by these technologies.
Disabled people in particular should be included from the beginning rather than consulted after systems have already been designed.
“Nothing about us without us” must apply to artificial intelligence too.
Anthropic Has Itself Warned Humanity About Losing Control
There is another extraordinary element to this story.
Anthropic itself has warned that the development of AI may eventually progress faster than society’s ability to control it.
In June 2026, the company called for major AI laboratories to consider mechanisms for a coordinated and verifiable development pause if certain danger thresholds were reached, warning about increasingly autonomous systems and the possibility of AI eventually contributing to the development of still more capable AI.
That does not prove that catastrophic AI scenarios will happen.
But when the people building some of the world’s most powerful systems are themselves discussing mechanisms to stop, pause or control development, governments and citizens should pay attention.
The Bigger Question: Who Watches the Watchers?
Anthropic says it needs safeguards.
The Pentagon says private companies should not dictate national-security policy.
Donald Trump says elected government must retain control.
Elon Musk has repeatedly warned that advanced AI requires oversight while simultaneously developing his own frontier AI systems.
- Other AI companies are racing to compete.
- Governments want strategic advantage.
- Investors want commercial returns.
- Military organisations want superior capabilities.
- Citizens want safety, liberty and privacy.
Every participant therefore has interests of their own.
That is precisely why no single participant should be trusted with unlimited power.
Where Does the Anthropic Controversy Stand Now?
As of 24 August 2026, the clearest legal position arising from the February confrontation is that President Trump’s government-wide directive and the Pentagon’s associated supply-chain measures were subjected to a preliminary federal injunction.
The injunction restored the earlier status quo while litigation continued.
The Trump administration subsequently denied Anthropic’s allegations of unlawful retaliation and continued contesting the company’s legal claims.
There have also been further tensions between the US government and Anthropic. In June 2026, access to certain newer Anthropic models was temporarily affected by US export-control measures before those restrictions were lifted on 30 June 2026.
This story is therefore far from a simple battle between a “good” company and a “bad” government, or vice versa.
It is an evolving struggle over power, constitutional authority, technology, national security and human rights.
Final Thoughts
Artificial intelligence may ultimately become one of humanity’s most beneficial inventions. It could help disabled people communicate, work remotely, navigate inaccessible environments, understand complex information, manage businesses and live more independently. It could accelerate medical research, improve assistive technology and give people abilities that previous generations could scarcely imagine. But the same underlying technology could also support surveillance, discrimination, cyberwarfare and increasingly autonomous weapons.
The Anthropic-Pentagon controversy should therefore serve as a warning.
The real question is not whether we support Anthropic, Donald Trump, Elon Musk, the Pentagon or any other individual organisation. The question is: “What safeguards should exist before anyone, company, government or machine, is entrusted with power over human rights, privacy and life itself”?
Artificial intelligence should remain a tool serving humanity, rather than humanity becoming subordinate to the systems it has created. For disabled and vulnerable people, that requires more than assurances from technology executives or politicians. It requires law, transparency, accessibility, independent oversight, enforceable human-rights protections and meaningful participation by the people whose lives will be affected. Because once AI becomes powerful enough to monitor populations, influence governments or participate in decisions concerning life and death, deciding who controls the technology may become one of the defining human-rights questions of our generation.
References and Further Reading
- https://www.anthropic.com/
- https://cand.uscourts.gov/judges/rfl/lin-rita-f
- https://www.ohchr.org/en/instruments-mechanisms/instruments/convention-rights-persons-disabilities
- https://www.facebook.com/share/v/1GFomZYdeM/
- https://docs.google.com/document/d/1mhYYAbtxZUm9h8CzjPEFzpyA-g35m-EDYCuao5L0nxQ/edit?usp=sharing
- https://www.anthropic.com/research/agentic-misalignment
- https://arxiv.org/abs/2510.05179
- https://time.com/article/2026/03/11/anthropic-claude-disruptive-company-pentagon/
- https://www.anthropic.com/news/statement-department-of-war
- https://www.reuters.com/technology/pentagon-informed-anthropic-it-is-supply-chain-risk-official-says-2026-03-05/
- https://www.anthropic.com/news/statement-comments-secretary-war
- https://www.politico.com/news/2026/03/05/pentagon-tells-anthropic-it-has-designated-the-company-a-supply-chain-risk-00814758
- https://edition.cnn.com/2026/03/05/tech/pentagon-anthropic-supply-chain-risk
- https://www.gsa.gov/about-gsa/newsroom/news-releases/gsa-issues-statement-on-anthropic-preliminary-injunction-04032026
- https://www.gsa.gov/about-gsa/newsroom/news-releases/gsa-issues-statement-on-anthropic-preliminary-injunction-04032026
- https://spectrumlocalnews.com/us/snplus/politics/2026/02/27/trump-orders-federal-agencies-phase-out-anthropic-ai-technology
- https://www.law360.com/articles/2461567/gsa-restores-anthropic-technology-post-injunction
- https://www.courtlistener.com/docket/72379655/anthropic-pbc-v-us-department-of-war/
- https://cases.justia.com/federal/district-courts/california/candce/3:2026cv01996/465515/135/0.pdf?ts=1774599805
- https://www.reuters.com/legal/litigation/trump-administration-denies-unlawful-retaliation-anthropic-ai-blacklisting-2026-06-09/
- https://digitallibrary.un.org/record/4080867
- https://social.desa.un.org/sdn/cosp18-event-artificial-intelligence-for-inclusion-strengthening-workforce-participation-for
- https://social.desa.un.org/issues/disability/crpd/article-31-statistics-and-data-collection
- https://www.reuters.com/legal/litigation/musk-proposes-peer-review-frontier-ai-models-economist-interview-2026-07-23/
- https://www.reuters.com/business/anthropic-says-ai-labs-need-coordinated-plan-halt-development-if-risks-rise-2026-06-04/
- https://www.wsj.com/tech/ai/anthropic-urges-global-pause-in-ai-development-flags-self-improvement-risk-99cefb73?msockid=16d952856a5f67960cbe45056bc866ab
Renata The Editor of DisabledEntrepreneur.uk - DisabilityUK.co.uk - DisabilityUK.org - CMJUK.com Online Journals, suffers From OCD, Cerebellar Atrophy & Rheumatoid Arthritis. She is an Entrepreneur & Published Author, she writes content on a range of topics, including politics, current affairs, health and business. She is an advocate for Mental Health, Human Rights & Disability Discrimination.
She has embarked on studying a Bachelor of Law Degree with the goal of being a human rights lawyer.
Whilst her disabilities can be challenging she has adapted her life around her health and documents her journey online.
Disabled Entrepreneur - Disability UK Online Journal Working in Conjunction With CMJUK.com Offers Digital Marketing, Content Writing, Website Creation, SEO, and Domain Brokering.
Disabled Entrepreneur - Disability UK is an open platform that invites contributors to write articles and serves as a dynamic marketplace where a diverse range of talents and offerings can converge. This platform acts as a collaborative space where individuals or businesses can share their expertise, creativity, and products with a broader audience.
