Anthropic Scales Back AI Safety Commitments Amid Competition & Pentagon Pressure
Anthropic, the AI company behind the Claude chatbot that was founded with a focus on safe technology, appears to be scaling back its safety commitments in order to keep the company competitive.
The company said on Tuesday it had changed its responsible scaling policy, a set of self-imposed guidelines aimed at preventing the development of AI that could potentially be dangerous and cause situations such as large-scale cyberattacks.
While the updated guidelines say Anthropic would still require a “strong argument that catastrophic risk is contained” when developing AI, they now state that development will only be delayed “until and unless we no longer believe we have a significant lead” — meaning the company would keep developing if it didn’t believe it have a lead over competitors.
The company said it has taken this step because concerns about the safety of AI in the U.S. have taken a back seat to its economic potential.
“Despite rapid advances in AI capabilities over the past three years, government action on AI safety has moved slowly,” the company said in a blog post.
“The policy environment has shifted toward prioritizing AI competitiveness and economic growth, while safety-oriented discussions have yet to gain meaningful traction at the federal level.”
The change in Anthropic’s safety guidelines comes as the Pentagon threatens to pull its contracts with the company unless its technology is allowed to be used for all legal military purposes — though Anthropic says the guideline change is unrelated.
The AI company has historically sold itself as putting safety first.
Anthropic was founded in 2021 by former employees of OpenAI who were concerned that company was putting development ahead of safety. CEO Dario Amodei has also voiced fears about the negative potential of AI, including mass human catastrophe, and maintained that safety continued to be the “highest-level focus” for Anthropic in a December interview with Fortune.
The blog post noted the company’s safety practices were always intended to be updated and that this new iteration improves the company’s “transparency and accountability” with new commitments to regularly publish reports and safety goals.
But Heidy Khlaaf, chief AI scientist at independent research group the AI Now Institute, says despite Anthropic’s safety-first reputation, it has always fallen short when it comes to its attempts to prevent human harm.
- Cross Country Checkup is asking: How are you using AI? Is AI making our lives better or worse? Leave your comment here and we may read it or call you back for Sunday’s show.
From its first safety policy, Khlaaf says, Anthropic has focused too much on the possibility of catastrophic events down the road, rather than the possibility of harm that could come from current AI technology, such as run-of-the-mill errors with chatbots.
The Claude chatbot has in the past been misused in fraud schemes and attempts to create malware and was recently used to steal Mexican government data, according to cybersecurity researchers.
She says the company is now dropping the “veneer of safety” it’s previously used to market itself because it’s become clear that’s not in its best interest.
“This is a strategic announcement to show that they’re open for business,” Khlaaf said.
Amid rapid global advances and deployment of artificial intelligence technologies, the federal government has invested millions to combine the minds of three existing institutes into one that can keep an eye on potential dangers ahead.
The announcement comes at a time of intense competition between top AI companies, such as Anthropic, OpenAI and Google, which have competing chatbots and are all striking deals to integrate their technologies with businesses and government departments.
U.S. President Donald Trump’s administration has also signalled it’s all-in on AI development and has threatened to withhold funding from states that enact laws that hold back U.S. dominance in the industry.
Teresa Scassa, Canada Research Chair in Information Law and Policy at the University of Ottawa, says that no-rules attitude from the U.S. government makes it hard for companies to prioritize safety, “because if they do that, then they are going to be left in the dust.”
That puts Canada in a tough place, too, she says, because regulation here could set homegrown AI development back compared to the U.S. or encourage Canadian companies to move south of the border, where there would be fewer restrictions on their tech.
“And I think the sense is that we can’t afford that in Canada right now,” Scassa said. “So you can see how it’s having that kind of knock on effect on AI regulation here.”
She says since Canada’s Artificial Intelligence and Data Act died in 2025, the Canadian government, much like its American counterpart, hasn’t tried to impose any broad AI regulation.
Safety change unrelated to Pentagon spat, company says
The change in Anthropic’s safety guidelines comes as it is under pressure from the Pentagon.
Anthropic struck a deal with the U.S. Department of Defence worth up to $200 million US in July, allowing the government to use its technology for military purposes but only within the company’s usage guidelines — the set of rules Anthropic has for how clients can and can’t use its products, including chatbot Claude.
An increasing number of people around the world are turning towards AI chatbots for companionship and connection. Meanwhile, a new study indicates those chatbots can even impact people’s voting intentions. Study co-author Gordon Pennycook, an associate professor at Cornell University in the U.S., discusses his research and how AI chatbots can affect potential voters differently than a traditional TV ad.
Those guidelines bar anyone, including the U.S. government, from using Anthropic’s AI tools for a range of things, including to design or develop weapons.
But according to reports, U.S. Defence Secretary Pete Hegseth issued CEO Amodei an ultimatum in a meeting on Tuesday — giving the company until Friday to allow the military to use its AI tools for all legal military purposes or risk losing its government contracts and being designated a supply chain risk.
Front Burner21:47Will AI agents take over the workplace?
In its back and forth with the government, Anthropic said it will not allow the government to use its technology in autonomous weapons systems — those that allow AI alone to fire at targets — and mass surveillance systems.
But Pentagon officials told media the dispute doesn’t involve AI’s potential uses in autonomous weaponry and mass surveillance and insist the government has always followed the law.
Anthropic says the update of its responsible scaling policy and demands by the Department of Defence are unrelated. Hegseth’s issues are with the company’s usage policy, rather than the scaling policy, according to Anthropic.
Ahead of the Friday deadline, Amodei said in a blog post that Anthropic would not accede to the administration’s wishes, underscoring the company’s opposition to use of its tech in domestic surveillance and autonomous weapons.
Amodei said he hoped the Pentagon would reconsider but that the company would “work to enable a smooth transition to another provider” if the Pentagon decided to cancel the contract.
“Our strong preference is to continue to serve the department and our warfighters — with our two requested safeguards in place,” Amodei wrote.
ARTICOL CREAT:
Anthropic Scales Back Safety Commitments in Competitive AI Landscape
Anthropic, the AI company founded on principles of safe and responsible technology development, is adjusting its core safety policies amid intensifying competition in the artificial intelligence sector. The shift, announced on Tuesday, signals a recalibration of priorities as economic pressures and a changing regulatory environment take hold. This move comes as the company navigates a complex landscape, facing both demands for broader access to its technology from the U.S. Department of Defense and scrutiny over its past safety record.
The company’s revised “responsible scaling policy,” a set of self-imposed guidelines designed to prevent the creation of potentially dangerous AI, now states that development will only be delayed “until and unless we no longer believe we have a significant lead” over competitors. Previously, the policy required a “strong argument that catastrophic risk is contained” before proceeding with AI development. The updated guidelines represent a notable departure from Anthropic’s earlier, more cautious stance.
A Shift in Priorities: Safety vs. Competitiveness
Anthropic’s decision reflects a broader trend within the AI industry, where the pursuit of economic growth and competitive advantage is increasingly overshadowing concerns about long-term safety. The company explicitly cited the slow pace of government action on AI safety in the United States as a key factor in its policy change. “Despite rapid advances in AI capabilities over the past three years, government action on AI safety has moved slowly,” Anthropic stated in a blog post. “The policy environment has shifted toward prioritizing AI competitiveness and economic growth, while safety-oriented discussions have yet to gain meaningful traction at the federal level.”
This shift is particularly notable given Anthropic’s origins. The company was established in 2021 by former employees of OpenAI who expressed concerns that their previous employer was prioritizing development speed over safety considerations. CEO Dario Amodei has consistently voiced anxieties about the potential negative consequences of AI, including the possibility of widespread catastrophe, and has maintained that safety remains Anthropic’s “highest-level focus” as recently as a December interview with Fortune magazine.
Pentagon Pressure and Usage Guidelines
The policy adjustment coincides with a separate, though Anthropic insists unrelated, dispute with the U.S. Department of Defense. In July, Anthropic secured a deal worth up to $200 million to allow the government to utilize its technology for military purposes, subject to the company’s existing usage guidelines. These guidelines prohibit the use of Anthropic’s AI tools for developing weapons or engaging in mass surveillance.
However, according to reports from the Wall Street Journal, Defense Secretary Pete Hegseth issued an ultimatum to Amodei, demanding that Anthropic permit the military to use its AI tools for all legal military applications or risk losing its government contracts and being designated a supply chain risk. Anthropic has reportedly resisted these demands, reaffirming its opposition to the use of its technology in autonomous weapons systems and domestic surveillance. Amodei stated in a blog post that the company would “work to enable a smooth transition to another provider” if the Pentagon decides to cancel the contract.
Past Misuse and Evolving Concerns
The change in Anthropic’s safety guidelines has prompted criticism from some experts who argue that the company has historically fallen short of its safety promises. Heidy Khlaaf, chief AI scientist at the AI Now Institute, contends that Anthropic has consistently focused on hypothetical, large-scale risks while overlooking the more immediate harms posed by current AI technology, such as errors and biases in chatbot interactions.
The Claude chatbot, developed by Anthropic, has been misused in fraud schemes and attempts to create malware. More recently, in February 2026, cybersecurity researchers discovered that the chatbot was used to steal sensitive data from the Mexican government. Khlaaf suggests that Anthropic is now shedding its “veneer of safety” as it becomes clear that prioritizing safety is not in its best commercial interest. “This is a strategic announcement to show that they’re open for business,” she said.
Canada’s Regulatory Landscape and the Broader Implications
The situation unfolding with Anthropic also highlights the challenges faced by countries like Canada in regulating the rapidly evolving AI landscape. Teresa Scassa, Canada Research Chair in Information Law and Policy at the University of Ottawa, notes that the lack of robust AI regulation in the United States creates a difficult environment for companies that prioritize safety. “If they do that, then they are going to be left in the dust,” Scassa explained. This dynamic could incentivize Canadian AI companies to relocate to the U.S. Or discourage investment in Canadian AI development.
Canada’s own attempt to establish a comprehensive AI regulatory framework, the Artificial Intelligence and Data Act, failed in 2025, leaving the country without a clear regulatory path forward. This lack of regulation, combined with the U.S.’s increasingly permissive approach, creates a challenging environment for Canada to foster responsible AI innovation.
What’s Confirmed and What Remains Unclear
Confirmed: Anthropic has revised its responsible scaling policy to prioritize competitiveness alongside safety. The company is facing pressure from the U.S. Department of Defense to broaden the permissible uses of its AI technology. Anthropic’s Claude chatbot has been subject to misuse in fraudulent activities and data breaches.
Unclear: The extent to which the Pentagon’s demands directly influenced Anthropic’s policy change remains disputed. The long-term impact of the revised policy on Anthropic’s safety practices and the development of its AI technology is yet to be seen. The future of Anthropic’s contract with the U.S. Department of Defense is uncertain.
Looking Ahead: Navigating a New Era of AI Development
The coming weeks will be critical for Anthropic as it navigates its dispute with the Pentagon and implements its revised safety policy. The company’s decisions will likely have broader implications for the AI industry, potentially setting a precedent for how companies balance safety concerns with competitive pressures. The situation also underscores the urgent need for clear and comprehensive AI regulation, both in the United States and internationally, to ensure that the benefits of AI are realized while mitigating its potential risks. The evolving dynamic between government, industry, and research institutions will be central to shaping the future of AI development and deployment.