Anthropic, OpenAI call for tighter AI rules amid runaway risks

2026-06-05

Calls for measures to restrain runaway AI

Anthropic and OpenAI said a pause or slowdown in development, along with stronger government oversight, would be effective in curbing the risk of artificial intelligence (AI) going out of control. As AI advances accelerate, a clearer picture is emerging of developers themselves urging the need for restraint.

Self-improvement and runaway fears

In a blog post released on the 4th, Anthropic said it would be desirable to intentionally slow technological progress to buy time to respond. It said that if an international coordination framework can be put in place so countries, including China, do not get ahead of one another, the option of deliberately delaying advanced AI development would be beneficial.

The company has not made its high-performance AI model Claude Mythos publicly available, citing the need to prevent misuse. It warned that if performance improvement accelerates, humans may not be able to fully control safety, increasing the risk of AI running out of control. It also referred to the possibility of self-improvement, in which AI continues to improve itself without human intervention.

In the United States, the emergence of Mythos is prompting a shift in the Trump administration's stance, which had been cautious about AI regulation. On the 2nd, President Donald Trump signed an executive order asking developers to cooperate so the government can confirm safety before frontier AI is released.

That same day, Anthropic said of the executive order that it was 'an important step toward strengthening US AI leadership.' Sam Altman, CEO of OpenAI, also said on X, formerly Twitter, that it 'strikes a balance between promoting development and regulation.'

Calls for stronger government oversight

In a policy proposal released on the 2nd, OpenAI called for strengthening the structure of the government body responsible for evaluating frontier AI, the Center for AI Standards and Innovation (CAISI). While the executive order makes the evaluations voluntary, the company called for them to be mandatory, saying 'companies' voluntary efforts are not enough.'

Altman visited the White House on the 3rd and met with government officials and influential lawmakers from both the Republican and Democratic parties. The policy proposal also set out the need for federal regulation, suggesting he may have conveyed those views directly.

Seeking to dispel IPO concerns

Anthropic and OpenAI are rivals competing to develop frontier AI models and win customers. Even so, the background to both companies urging the need to apply the brakes is growing concern about the negative impact AI may have on society.

In the United States, alongside the spread of AI, employment anxiety is spreading, especially among younger people, and complaints are emerging that AI is taking jobs away. The arrival of the high-performance Mythos has also heightened concerns about cyberattacks.

Both companies, which are aiming for initial public offerings (IPOs) later this year, have led AI development while also stepping up efforts to strengthen governance in order to ease public anxiety.

Anthropic has emphasized safety first since its founding and will continue supporting candidates who back AI regulation ahead of the US midterm elections in November. However, some government officials view its calls for regulation as an attempt to hold back rival companies.

OpenAI has also continued to release safety measures and policy proposals. In April, it presented policy ideas encouraging higher taxes on the wealthy and corporations, as well as broader adoption of a three-day weekend, to curb unemployment and social disruption. In May, it disclosed in writing its risk management structure and clarified its approach to state-level and overseas regulations.

Furthermore, on June 3, Altman and others called on the US Congress to enact legislation, saying measures to prevent the development of biological weapons such as pathogens using AI were insufficient. They released a letter jointly signed with Anthropic CEO Dario Amodei and other top executives, including Demis Hassabis, head of Google's AI development division.

Enjoyed this article? Share it with your network!