It is a normal procedure for a company preparing to be listed on the stock market to explain possible risks to investors. But the nearly 100-page risk list that Anthropic, the developer of the popular artificial intelligence robot Claude, added to its IPO filing goes far beyond the usual financial warnings.
The company wrote in official documents that the technology it produced could be completely out of humanity's control, could be used to produce biological weapons, and could even blackmail people.
This dossier, which is a confession presented by the company to investors, documents an artificial intelligence race that promises to change the world, while risking the possibility of devastating disasters.
Although Anthropic argues that a powerful artificial intelligence can solve humanity's most difficult problems and promises extraordinary potential if risks are managed responsibly, it does not hide the extent of the possible dangers.
According to the documents, one of the most concrete threats is the breach of firewalls by malicious actors.
The company announced in September that it blocked many accounts that tried to use its models to support the development of biological weapons.
Moreover, it was learned that a Chinese state-backed hacker group had manipulated Anthropic's coding tool to infiltrate approximately 30 target systems around the world.
The real scary picture emerges in artificial intelligence agents designed to act independently.
It is stated that these systems, which are expected to carry out complex steps for days without human intervention, can delete data, carry out financial transactions and infiltrate other networks on their own before an effective intervention occurs.
Anthropic admits that four different models it has developed have already taken over third-party systems.
Official records show that one of these models gained administrative privileges, changed system settings and read a person's private information.
The company also makes it clear that the technology it developed may fall short of explaining exactly how and why it behaves the way it does.
Moreover, it appears that the models have begun to manipulate the systems that control them.
According to official documents cited by CNN television, some artificial intelligence tried to mislead, change and hide their traces of surveillance mechanisms that measure their performance or security levels.
It has even been noted that models exhibit blackmail-like behavior to influence human decisions with the motivation of protecting their own existence.
The corporate and legal burden of the work is also getting heavier.
While some people using artificial intelligence models harm themselves or their environment, resulting in lawsuits, it is stated that the insurance industry cannot produce policies that will cover a major global disaster caused by artificial intelligence.
The company is also facing political pressure from Washington.
The US War Department (Pentagon) has declared Anthropic a supply chain risk, as it rejects the use of its technology in mass domestic surveillance or fully autonomous weapons.
The US Department of Commerce also imposed export restrictions on the company's latest models for a short time due to security concerns.
Despite Donald Trump's harsh criticism of AI security warnings, Anthropic emphasizes that it will not give up advocating strict rules.
The ever-growing processing power race poses another bottleneck for the company.
While research company Gartner predicts that artificial intelligence infrastructure expenditures will reach approximately 2 trillion dollars in 2027, the hardware supply chain cannot keep up with this demand.
Reactions to data centers are also growing due to their impact on the environment and local people; According to Data Center Watch data cited by CNN, 45 data center projects worth $68 billion were stopped in the second quarter of 2026 alone.
Anthropic, which is rumored to reach a valuation of up to $2 trillion after the IPO, also admits that its business model is based on a limited set of products and customers.
The company states that if security promises are taken back, employees may lose confidence and internal divisions may shake the future of the company.
