Former AI Researcher Testifies as Pentagon Cuts Ties With Anthropic Over Safety Dispute
A former AI researcher at Anthropic and OpenAI, Jacob Coxon, is set to testify at a New York City hearing about the risks of advanced AI systems. His appearance comes as the Pentagon halts the use of Anthropic’s AI tools after designating the company a "national security supply-chain risk." The dispute centers on Anthropic’s refusal to remove safety guardrails for military applications, raising broader questions about AI ethics and government oversight.
Editor, Lazyfounder

30 SEC SUMMARY
- Jacob Coxon, a former AI researcher at Anthropic and OpenAI, will testify at a New York City AI hearing about risks posed by advanced AI systems.
- Anthropic has been designated a "national security supply-chain risk" by the Pentagon, leading to a legal dispute over its AI tools' military use.
- The Pentagon halted Anthropic’s AI tools after the company refused to remove safety guardrails, citing concerns over surveillance and autonomous weapons.
- Anthropic’s CEO Dario Amodei faces criticism from a Republican senator for warning about catastrophic AI risks ahead of the company’s expected IPO.
- Anthropic updated its usage policy to ban abusive interactions, election interference, and weapons-related applications of its AI models.
TABLE OF CONTENTS
- Former AI Researcher to Testify at New York City Hearing
- Pentagon Designates Anthropic a Supply Chain Risk, Halts AI Tools
- Broader Tensions Between Tech and Government Over AI Control
- Anthropic CEO Criticized for Alarmist AI Warnings Ahead of IPO
- Background: The Pentagon-Silicon Valley Dynamic
- What this means
- Key takeaways
- Sources
KEY HIGHLIGHTS
- Jacob Coxon, a former researcher at Anthropic and OpenAI, will testify at a New York City AI hearing at the request of Council Speaker Julie Menin.
- Coxon has warned that AI could "kill us all by the end of the decade" and accused AI companies of "gambling with our lives."
- The Pentagon designated Anthropic a "national security supply-chain risk" in February 2026 and ceased using its AI tools after the company refused to remove safety guardrails.
- Anthropic’s Claude model was deeply integrated into the Pentagon’s Maven Smart System, used for intelligence gathering and military operations.
- Republican Senator Bernie Moreno criticized Anthropic CEO Dario Amodei for his "alarmist approach" to AI risks ahead of the company’s planned IPO.
- Anthropic updated its usage policy to prohibit prolonged verbal abuse, election interference, weapons software, and surveillance.
Former AI Researcher to Testify at New York City Hearing
Jacob Coxon, a former AI researcher at Anthropic and OpenAI, will testify at a New York City Council hearing on artificial intelligence on October 5, 2026. The hearing was organized at the request of Council Speaker Julie Menin, who aims to establish safety rules for AI technology.
Coxon has publicly warned that AI could "kill us all by the end of the decade" and accused companies like Anthropic and OpenAI of "gambling with our lives," according to reports from Mint (Technology).
Coxon, a silver and bronze medalist at the International Mathematical Olympiad, resigned from Anthropic last month after expressing concerns about the company’s commitment to AI safety. He compared building superintelligent AI to "summoning an alien species," arguing that researchers cannot fully control such systems.
The hearing will also include executives from OpenAI, Anthropic, and Alphabet Inc.’s Google, per Mint (Technology).
Pentagon Designates Anthropic a Supply Chain Risk, Halts AI Tools
The Pentagon has stopped using Anthropic’s AI tools after designating the company a "national security supply-chain risk" in February 2026, according to BBC News (Tech & Business). The designation, typically reserved for companies in adversarial countries, was upheld by a federal court.
The Department of Defense had set a deadline of late August 2026 to phase out Anthropic’s tools, including its Claude model, which was still in use as recently as last week. Claude was integrated into the Maven Smart System, a Pentagon platform for organizing intelligence and data operated by Palantir.
The Pentagon pressed Anthropic to remove safety guardrails from Claude to enable unfettered military use, but the company refused, citing concerns over mass surveillance and autonomous weapons. Anthropic has since sued the Trump administration to overturn the designation, calling it "unprecedented and unlawful."
In place of Anthropic, the Pentagon has signed contracts with Google, xAI, and OpenAI, with OpenAI’s tools being adopted more widely in recent months, per BBC News (Tech & Business).
Broader Tensions Between Tech and Government Over AI Control
The dispute over Anthropic reflects a growing conflict between the U.S. government and private tech companies over control of AI technology in military and national security applications. Author Sharon Weinberger, interviewed by Gizmodo, described this dynamic as the rise of a "Pentagon-Silicon Complex," where tech CEOs wield significant influence over national security policies.
Tech companies like Palantir and Anduril have positioned themselves as key players in defense contracting, often retaining ownership of their technology rather than ceding control to the government. Anduril’s drones, for example, faced technical glitches in Ukraine, raising concerns about reliability and accountability.
Palmer Luckey, founder of Anduril, has openly expressed willingness to build nuclear weapons, citing their historical role in geopolitical stability. Meanwhile, Palantir’s valuation of $455 billion is driven by narratives about AI, despite the company not training its own AI models.
Anthropic CEO Criticized for Alarmist AI Warnings Ahead of IPO
Republican Senator Bernie Moreno has accused Anthropic CEO Dario Amodei of taking an "alarmist approach" to AI risks, questioning how the company can seek public investment while warning of existential threats. Moreno’s letter, sent ahead of Anthropic’s expected IPO, cited concerns raised by researchers like Coxon and Evan Hubinger, who estimate a greater than 10% chance of catastrophic AI outcomes within a decade.
Investors are reportedly discussing a valuation of nearly $2 trillion for Anthropic, roughly 42 times its revenue. The company’s IPO prospectus has already warned of risks from U.S. government action, according to The Next Web.
Anthropic recently updated its usage policy to explicitly prohibit prolonged verbal abuse of its models, election interference, weapons software, and surveillance. The policy also bans deceptive campaigns, such as using Claude to run fake accounts or generate fabricated news.
Background: The Pentagon-Silicon Valley Dynamic
The relationship between the U.S. government and tech companies has evolved significantly over the past decade, with private sector leaders like Elon Musk, Alex Karp, and Palmer Luckey positioning themselves as pivotal figures in national security. This shift mirrors the post-World War II "military-industrial complex," but with Silicon Valley increasingly dictating the terms of engagement.
Historically, defense contracts were dominated by traditional aerospace and military contractors like Lockheed Martin and Northrop Grumman. However, the rise of AI and autonomous systems has created opportunities for tech startups like Anduril and Palantir to challenge established players, often with less oversight.
What this means
Lazyfounder analysis — our interpretation, not reported fact.
This story underscores a critical inflection point for the AI industry, where ethical commitments, government demands, and commercial interests collide. For founders and operators, Anthropic’s situation is a cautionary tale about the risks of aligning—or refusing to align—with government priorities, particularly in high-stakes sectors like defense and national security.
The Pentagon’s move to cut ties with Anthropic over safety guardrails signals that regulators and military agencies are no longer willing to defer to tech companies’ self-policing. This could accelerate efforts to standardize AI safety protocols, but it also risks stifling innovation if compliance becomes overly burdensome.
For startups in AI or adjacent fields, the broader takeaway is the growing scrutiny over how technology is deployed, especially when it intersects with public policy or national security. Companies must weigh the PR and regulatory risks of working with governments against the financial and strategic benefits—and prepare for pushback from employees, investors, or lawmakers if their stance shifts.
Finally, Anthropic’s IPO plans highlight the tension between warning of existential risks and seeking public investment. Founders should expect investors and regulators to demand clarity on how such warnings align with business models, particularly in industries where profitability depends on scaling rapidly.
Key takeaways
- AI safety debates are intensifying as researchers like Coxon publicly challenge industry practices, urging stricter oversight and transparency.
- Government agencies and tech companies are clashing over control of AI tools, particularly in military applications, highlighting broader conflicts over ethical guardrails.
- Anthropic’s legal and regulatory challenges could set precedents for how AI companies navigate government demands and public accountability.
- Public testimony and policy updates from AI companies reflect growing pressure to address misuse, election interference, and national security risks.
- Investors and regulators are closely watching Anthropic’s IPO amid controversies over AI risks, military contracts, and ethical commitments.
Related on Lazyfounder
Sources
- Mint (Technology) · 2026-10-05
Who is Jacob Coxon? Former Anthropic researcher to testify at New York City AI hearing today - Gizmodo · 2026-10-05
Beware the Rise of the ‘Pentagon-Silicon Complex’ - BBC News (Tech & Business) · 2026-10-05
Pentagon stops using Anthropic AI tools months after blacklisting company, BBC told - The Next Web · 2026-10-05
Senator Bernie Moreno accuses Anthropic’s Amodei of ‘alarmist’ AI talk - TechCrunch · 2026-10-08
Anthropic changes usage policy to ban model abuse and election interference
This story is an original summary drafted with AI by Lazyfounder from the reporting listed above and checked by automated validation. Facts are attributed to their original publishers; sections marked as analysis are Lazyfounder's. Where a source is in another language, facts were machine-translated and quotations are reported, not reproduced. Read the original coverage via the links, and see our AI policy and corrections policy.
About the author
Editor, Lazyfounder
Tarun Mottlia edits LazyFounders, covering Indian startups, funding rounds, AI and product launches. Every story on the site is AI-assisted and checked against its cited sources before publication.
More stories by Tarun MottliaGet the LazyFounder Brief
Startup, funding and AI news in a five-minute read. Join the early-access list.


