Anthropic has built its public identity around AI safety, while also becoming one of the companies most actively advancing frontier artificial intelligence. That tension is now central to how observers understand the company: it warns that advanced AI could produce grave harms, yet believes it must remain near the front of the race to reduce those risks.
The Argument Behind Anthropic’s Growth
According to several former employees who spoke to WIRED on the condition of anonymity, Anthropic is guided by two main beliefs. The first is that artificial intelligence will be extraordinarily transformative and that its arrival is not something the company can stop. The second is that the outcome will be better if Anthropic remains one of the leading AI developers.
That view helps explain why the company can speak so starkly about danger while pursuing more capital, compute, research talent and political influence. Inside the company, those resources are not described as separate from the safety mission. They are treated as tools needed to shape the future of AI from a position of strength.
Anthropic’s mission is “to ensure the world safely makes the transition through transformative AI.” The company’s internal logic is that it cannot fulfill that mission from the sidelines. To influence standards, safeguards and deployment choices, it believes it must be technically competitive with the most advanced AI labs.
That has made Anthropic both a warning voice and a major market actor. The company is among the top developers and distributors of cutting-edge AI models, courts customers like the US military, and was recently valued at almost $1 trillion.
Safety From Inside the Race
Helen Toner, executive director of Georgetown’s Center for Security and Emerging Technology and a former OpenAI board member, described Anthropic’s worldview through an analogy about a forest containing both valuable treasures and dangerous monsters. In that telling, many groups are moving toward the rewards, and Anthropic wants to go farther while investing heavily in controlling the hazards.
Toner told WIRED that Anthropic’s strategy is distinctive because the company believes others will move ahead regardless. Its answer is to build at the frontier, become credible, and use that position to argue for safeguards. She said the company is direct about this approach, even if it can sound unusual from the outside.
Anthropic CEO Dario Amodei has described the need to combine competitiveness with safety. In a conversation with his cofounders posted on the company’s career page, he said: “You have to find a way to actually be competitive, to actually lead the industry in some cases, and yet manage to do things safely.”
That sentence captures the company’s core tradeoff. Anthropic does not present power as a distraction from safety. It presents power as the condition that makes safety work possible at the level it thinks matters.
A Company Built Against a Cautionary Example
Anthropic was founded in 2021 by former OpenAI employees who had lost confidence in OpenAI leadership, particularly CEO Sam Altman, to safely bring transformational AI into the world. WIRED reports that this origin story still shapes the company’s self-understanding.
Two former employees said Anthropic executives often refer internally to Altman and OpenAI, and to a lesser extent Meta and Elon Musk’s xAI, as cautionary examples. In that framing, Anthropic sees itself as a more responsible steward of powerful AI systems.
Former employees also said the company’s mission focus is unusually intense. One said Anthropic tells job candidates it is not a typical company driven only by market forces. The company points to a public benefit structure that allows it to prioritize the “long-term benefit of humanity” above profits.
At the same time, Anthropic sees commercial success and powerful AI models as part of that responsibility. The company’s position is not that profit and scale are irrelevant. It is that those things can serve the broader obligation to lead the industry on safety.
Sam McCandlish, cofounder and chief architect of Anthropic, put the founding motivation bluntly in the same company career-page conversation: “None of us wanted to found a company, we just felt like it was our duty.” Anthropic declined to comment for WIRED’s story.
The Accountability Problem
The unresolved question is whether a company can safely accumulate that much influence while also judging itself to be the right organization to hold it. Anthropic says on its website that it is a “high-trust, low-ego organization.” Former employees told WIRED that this description is largely accurate.
They also said employees generally have more faith in Amodei than they do in leaders at other AI labs. That trust extends to his statements about technological progress, government interactions and geopolitics.
But trust can create its own risks. Shazeda Ahmed, a postdoctoral scholar at UCLA who has studied the ideological origins of the AI safety movement, told WIRED that groups like Anthropic can struggle with a lack of pluralism. Her research found that the AI safety movement, rooted in subcultures including effective altruism, often suffers from homogeneity of thought and tends toward self-governance.
Ahmed’s concern is that people who share the same assumptions may not challenge one another deeply enough. In her words: “You’re not being challenged on these ideas when you surround yourself with other people who believe them.”
Former employees offered different pictures of internal debate. One described a lively culture where staff criticism can lead to lengthy responses from leadership. Another said more candid criticism often stayed in private group chats and did not become direct challenges to Amodei’s decisions.
The Palantir Deal Shows the Tension
One major internal controversy came in the fall of 2024, when Anthropic became the first AI lab to partner with Palantir to provide AI services to US intelligence and defense agencies. Some former employees said questions were raised inside the company, but those debates did not change Anthropic’s policies.
Anthropic employee Evan Hubinger wrote on LessWrong that the company was “extremely forthright” with staff about the Palantir deal. He also wrote that while some lines probably should not be crossed without careful consideration, the deal was overall a positive development.
The episode illustrates the broader dilemma. Anthropic believes frontier AI safety requires a seat at the table, and that seat comes from building, selling and influencing powerful technology. Critics and former employees are asking whether the same belief can make it harder for the company to see its own blind spots.
For Anthropic, the central claim is clear: the safest path through transformative AI is not retreat, but leadership. The unresolved issue is who gets to decide when leadership becomes too much power.