Three Key Areas Where Policymakers Can Support AI, According to Anthropic's CEO

Dario Amodei, CEO of Anthropic, discussed how U.S. policymakers can contribute to the development of artificial intelligence. Speaking at the Anthropic Futures Forum, he emphasized the need to balance innovation with safety. According to him, there are three main areas the government should focus on: export controls, basic safeguards, and support for job losses. These points are based on his experience developing large language models and agentic artificial intelligence tools. Amodei warned that artificial intelligence represents both the greatest threat and the greatest opportunity for national security in the past hundred years.

In an interview conducted by reporter Alexandra Kelley, Amodei described how his company approaches the safe deployment of artificial intelligence. He cited examples of its use in medicine and scientific research while also acknowledging the potential risks of advanced systems. According to him, the government has a key role to play in managing these risks. Amodei plans to meet with members of Congress, along with Anthropic co-founder Jack Clark, to discuss these ideas directly on Capitol Hill.

Export Controls as a Priority

Amodei emphasized the importance of export controls to prevent hostile actors from misusing advanced artificial intelligence technologies. His company, Anthropic, recently restricted access to its models for Chinese companies and their subsidiaries. This move cost the company hundreds of millions of dollars in revenue—at the current exchange rate, that amounts to approximately CZK 2.25 billion to CZK 20.25 billion. Amodei proposes applying similar restrictions to semiconductor chips, which are a key component in building these systems.

According to Amodei, the technology used to manufacture these chips is highly complex, and the U.S. has an advantage in this area. The company consistently advocates these measures at every level of the technology supply chain, not only for artificial intelligence models. This approach is not about manipulating the market but about protecting against misuse. Other sources confirm that Amodei is calling for restrictions on the spread of advanced models and hardware to hostile countries because of national security risks. He also estimates a 25 percent chance of catastrophic risks if these threats are not addressed, including misuse in cyberwarfare or biological engineering.

Basic Safeguards

Another area is the introduction of basic safeguards to ensure corporate accountability. Amodei supports transparency requirements for training artificial intelligence models and the implementation of basic safety tests whose results would be publicly available. These safeguards should be minimal so as not to hinder innovation and the benefits of artificial intelligence.

Amodei agrees with Trump administration officials that policy should be flexible so as not to restrict progress. For example, companies should share information about model training, which would help ensure that the technologies remain safe. Related information from online sources adds that Amodei advocates international cooperation on regulations that would monitor catastrophic risks. This includes public safety tests and minimum requirements that companies would have to meet without slowing development.

Support for Job Losses

Amodei acknowledges that artificial intelligence could cause massive job losses, particularly in white-collar occupations. According to him, this could affect 300 million Americans and many people in other countries. There is no perfect solution, but it is necessary to discuss the issue openly and prepare for it.

He proposes public awareness campaigns, better education for lawmakers, and debates about solutions such as retraining programs or the equitable distribution of wealth generated by artificial intelligence. Amodei mentions the possibility of taxes on artificial intelligence companies to fund these transitions. Other sources state that Amodei warns of a rapid rise in unemployment and calls for early debates about social safety nets and economic adaptation. He emphasizes the need for joint committees and briefings for lawmakers to help them better understand the impacts of artificial intelligence.

Source: nextgov.com