OpenAI's Altman said the company may slow down the pace of artificial intelligence development as r

OpenAI internal meeting revealed that it may proactively slow down the pace of cutting-edge AI research and development

This week, OpenAI employees learned that this artificial intelligence laboratory plans to proactively slow down the pace of work in the most advanced field of artificial intelligence. CEO Sam Altman also made the offer to competitors, including Anthropic. Anthropic acknowledged the risks associated with AI development in a report that revealed attempts by scientists connected to government agencies to use its Claude model for biological weapons research.

OpenAI said that due to security concerns, the company has suspended some internal training runs and relaxed the development of some models. The successive emergence of these disclosures has intensified debate over whether the industry is moving too fast ahead of the public offering, as the two companies continue to test the market.

Why does Altman want to slow down AI development?

According to sources cited by Bloomberg, at the company's plenary meeting, OpenAI CEO Sam Altman told attendees that the laboratory and other volunteer laboratories could slow down the release of cutting-edge models. Due to the voluntary nature of such arrangements, Altman acknowledged that not all companies would support the plan.

There are precedents for this skepticism. Three years ago, in March 2023, more than 1000 people including Elon Musk and Steve Wozniak signed an open letter calling for a six-month suspension of advanced AI work. At the time, media reported that the laboratories had not truly fulfilled the ban they jointly reached.

In addition, Jakub Pachoki, chief scientist of OpenAI, also issued an AI risk warning in a post posted on September 6, calling on AI companies to coordinate "slowing future development as needed." Pachoki said he hoped voluntary slowdowns would "become the norm before shared security standards were established."

In July, Altman said he had raised the need to slow AI development to White House officials. Previously, he and Anthropic CEO Dario Amodei signed a brief statement defining AI as a social-scale extinction risk.

Can AI models be used to make biological weapons?

The development of AI models has reached a stage where they may support biological weapons research. Just this week, OpenAI announced that its model had solved the Navier Stokes mathematical problem, although the result was controversial.

Anthropic's 154-page threat intelligence report released on September 10 provides new input to the debate. The report details five cases in which researchers tried to use Claude's capabilities to support his biological weapons work. Anthropic said attempts from "unsupported areas" tried to circumvent geographical limitations and obscure their research purposes. The laboratory said it had closed the relevant accounts, but did not specify the specific country or institution.

Do the risks of AI outweigh the benefits?

The reports were released at a time when public panic was spreading within the laboratory. On September 8, researcher Jacob Coxon resigned and accused OpenAI and Anthropic of "risking our lives." He said colleagues believed the technology could "kill us all by the end of the decade" and warned that humans could face extinction by 2030. His posts have received more than 150 million views.

Two days later, on September 10, former employees from Anthropic and Google DeepMind expressed similar concerns and called for greater transparency. A petition in late July to establish mechanisms to slow development collected signatures from more than 1000 employees at major AI companies.

Currently, OpenAI is pushing Congress to create mandatory, capability-based rules covering areas such as testing, independent evaluation, cybersecurity and incident reporting. "The prospects of AI accelerating AI development require measures that go beyond voluntary commitments," wrote Chris Lehane, chief officer of global affairs.

Policy-level response

California takes the lead: Governor Gavin Newsom signed into law on Wednesday SB 813 and AB 1405, as well as a supporting bill AB 1864 aimed at preventing biological threats caused by AI.

In Washington, D.C., Senator Bernie Sanders introduced legislation that proposes a moratorium on advanced AI development and a ban on artificial superintelligence. He told the BBC that ignoring scientists 'warnings would make anyone look "stupid".

President Donald Trump countered, expressing concern on Thursday that "if we don't win in AI, we will be at a very disadvantage."