Anthropic’s AI Ambition Raises a Bigger Question: Who Controls Its Power?

Anthropic CEO Dario Amodei has offered an optimistic vision for artificial intelligence. He believes powerful AI could make the world “stunningly beautiful” if developed in the right direction. His vision includes cures for many diseases, longer human lifespans and an end to poverty.

Yet Anthropic faces a difficult question as its technology advances. How much power should an AI company hold over systems that could reshape society?

The debate has moved beyond AI productivity. It now focuses on who decides what AI systems can access, what they can do and where their authority ends.

Anthropic’s Claude models have become major competitors in the frontier AI market. Its Claude Code tool has also gained significant attention among software developers.

Recent industry data cited in discussions suggests that 39% of developers worldwide used Claude Code between May and July. The figure reportedly reached 47% among US developers.

The company’s growing market position has also increased investor interest. Reports suggest Anthropic could pursue one of the largest technology IPOs in history. The company has reportedly discussed a potential addressable market exceeding $30 trillion.

However, commercial success has brought greater scrutiny. Critics are asking whether a company created to build safer AI could itself become too influential.

From AI idealism to growing influence

Anthropic was founded in 2021 by former OpenAI employees, including Amodei. The founders wanted to develop advanced AI with stronger safety measures.

That mission remains central to the company’s identity. However, frontier AI development requires huge amounts of computing power, data and capital.

This creates a difficult tension. More capable AI systems need broader access to information and greater ability to act. Companies also have strong commercial incentives to expand those capabilities.

Azhar Ali Blames PCB Instability for Pakistan Cricket’s Decline

Claude Code has become part of this debate.

Developers have raised questions on GitHub and technology forums about the tool’s network activity. Some discussions have focused on network requests and information about websites or domains during operation.

Those reports do not prove covert surveillance by Anthropic. However, they have increased discussion around telemetry, transparency and user consent.

The debate became more serious around Claude Code’s permission controls. Anthropic offers a setting called “dangerously-skip-permissions”. It allows users to bypass confirmation prompts and let the system execute operations directly.

The company has also introduced an auto mode. A built-in risk classifier determines whether certain operations require additional permission.

These features raise a wider issue. When does an AI assistant stop simply following instructions and start making consequential decisions for its user?

The growing problem of AI authority

Anthropic says it wants to build powerful AI while maintaining safety controls.

Critics, however, argue that greater autonomy inevitably changes the balance of control.

The progression is simple. First, AI systems observe more information. Then they gain permission to act. Eventually, they can decide whether an action appears safe enough to complete without asking.

Supporters see this autonomy as essential for useful AI agents. Critics fear that users may not fully understand the authority they are giving these systems.

Therefore, the debate is no longer only about technical accuracy.

It is increasingly about governance.

Who creates the rules? Who defines acceptable risk? Who can change those rules? Most importantly, who remains accountable when an autonomous AI system makes a consequential decision?

Anthropic’s role in national security

Amodei’s influence also extends beyond Anthropic’s commercial products.

In an earlier essay on AI and international security, he argued that democratic countries should cooperate on critical parts of the AI supply chain. He also supported restrictions on adversaries’ access to advanced chips and semiconductor equipment.

Amodei has additionally argued that AI could strengthen military capabilities. He has called for closer cooperation between governments and private AI companies.

That position reflects a wider shift in Silicon Valley.

AI companies are no longer focused only on consumer products. Their technology is becoming increasingly important to national-security strategies.

Palantir CEO Alex Karp has made a similar argument in his book The Technological Republic. He advocates a stronger role for technology companies in national defence and governance.

As technology and government become more closely connected, corporate power is also expanding.

Frontier AI companies now combine advanced technology, enormous investment, large workforces and growing influence over national-security policy.

That creates a difficult question. Can companies developing some of the world’s most powerful technologies remain purely commercial organisations?

Anthropic and Washington face a test of authority

Anthropic’s relationship with the US government highlights this tension.

Claude has reportedly been used in government and defence-related applications. These include intelligence analysis, modelling, simulation, operational planning and cyber-related work.

However, the relationship has also exposed disagreements about government control.

Government agencies argue that AI systems used for national security must ultimately support government objectives.

AI companies, meanwhile, want certain safety restrictions to remain under their control.

The disagreement points to a much bigger challenge.

Who controls a frontier AI system when it becomes strategically important?

The company that developed it could claim authority. The government using it could make the same argument. Regulators could also seek control through laws and safety standards.

No universally accepted answer exists yet.

The growing AI lobbying battle

The political influence of AI companies is also expanding.

Anthropic reportedly increased its federal lobbying expenditure during the first half of 2026. The company has also committed tens of millions of dollars to initiatives focused on AI-risk regulation.

Critics see a potential conflict.

AI companies lobby policymakers about rules governing the technology. At the same time, those companies help define the risks that the rules are designed to address.

Supporters offer a different argument. They say frontier AI companies possess technical knowledge that governments urgently need.

Therefore, they argue, these companies should have a role in policy discussions.

The disagreement highlights a larger struggle over who gets to define AI safety.

China changes the strategic equation

The debate is unfolding alongside another major development.

The technological gap between leading Chinese and American AI models is narrowing.

Stanford’s 2026 AI Index has reported significant progress by Chinese models compared with their US counterparts. Chinese companies are also competing aggressively on cost.

Open-source and lower-cost models could make advanced AI more accessible. They could also make it easier for businesses to switch between competing systems.

Reports that major US technology companies have explored bringing Chinese AI models onto American cloud platforms further illustrate the changing market.

For years, US dominance rested partly on the belief that the most advanced AI systems would remain concentrated in American companies.

That assumption now faces greater pressure.

As model performance converges, restrictions on chips and computing infrastructure may become harder to use as permanent barriers.

The competition is therefore expanding beyond the models themselves.

It now includes chips, cloud infrastructure, data, applications, technical standards and regulation.

Who gets to define the future of AI?

Amodei’s vision remains ambitious.

Powerful AI could accelerate scientific discovery, improve medicine and increase productivity. It could also help humanity tackle problems that have resisted generations of human effort.

However, the same technology could concentrate enormous power among a small group of companies and governments.

That contradiction is becoming harder to ignore.

AI companies argue that they are working to make powerful technology safe for humanity. Critics respond that safety cannot depend entirely on the companies developing the systems.

The deeper challenge may therefore not be whether AI becomes powerful.

It almost certainly will.

The more important question is who decides how that power is used.

Anthropic may believe it is helping create a better future. Its critics fear that the effort to shape that future could give the company too much influence over the present.

Meanwhile, competition between American and Chinese AI companies is turning the issue into something much larger than a Silicon Valley debate.

It is becoming a question of national power, economic competition and global governance.

The future of AI could indeed become “stunningly beautiful”. But deciding who controls the road to that future may prove just as important as the destination.