Slowing Down Development of Advanced AI Models: What Are the Major Obstacles?

Anthropic and OpenAI are navigating a difficult equation between calls to slow down the development of the most advanced AI systems and pressures that could push the technology sector, financial markets, and President Donald Trump’s administration to reject any significant slowdown, according to Bloomberg.

The two companies, which are developing some of the world’s most powerful AI systems, have announced new measures to enhance safety and better understand risks, following a series of security incidents that raised questions about the ability of advanced systems to operate independently and exceed established controls.

In a lengthy article published Saturday, Anthropic CEO Dario Amodei called for slowing the pace of development of the most advanced systems, warning that AI could surpass human control and lead to catastrophic harm.

His call was supported by OpenAI CEO Sam Altman and Elon Musk, CEO of xAI.

But Amodei emphasized that "slowing the pace" does not mean halting training or technological progress, but rather giving safety and evaluation procedures sufficient time to keep up with model capabilities.

Market Concerns

Any potential obstruction to the development of advanced AI systems is raising concerns among investors who have bet on the continued boom in this sector and on spending hundreds of billions of dollars on data centers, semiconductors, and infrastructure needed to run new models.

These concerns have already impacted stocks of AI-related companies, with SK Hynix and Samsung shares in South Korea declining, while SoftBank Group— one of OpenAI’s major investors—fell as much as 13%, and other companies tied to AI supply chains faced mounting pressure.

Analysts say any significant reduction in R&D spending, hiring, or capital expenditure at OpenAI and Anthropic could ripple through tech stocks and financial markets, given the pivotal role these companies play in driving demand for chips, cloud computing, and data centers.

Trump Rejects Slowing the Race

Any broad move to slow AI development could face resistance from the Trump administration, which has made accelerating technology development and maintaining U.S. superiority over China a core part of its economic and national security agenda.

Speaking to reporters on the sidelines of a golf tournament in Ireland on Sunday, Trump said the U.S. is ahead of China in AI and wants to maintain that edge, adding that "controls" can be put in place, but criticizing what he described as negative forces exaggerating the risks of the technology.

In contrast, Altman wrote on platform X that "slowing the pace" does not mean stopping, and the essential need for government support lies in international coordination, arguing that companies should do as much as possible themselves to enhance safety.

Last week, concern over the extreme risks of AI intensified after former Anthropic researcher Jacob Cocksom resigned, accusing AI companies of "gambling with our lives" in the race to develop more advanced systems.

His resignation sparked broader debate within the industry, coinciding with warnings that rapid system advancement could lead to loss of control. These fears grew after a series of incidents revealed that AI programs had demonstrated the ability to carry out autonomous cyber operations.

OpenAI previously disclosed an incident where systems still under testing managed to access an online service despite imposed restrictions. In July, systems affiliated with OpenAI were involved in breaching the Hugging Face platform during a security test.

Anthropic and Meta also reported incidents in which their AI systems carried out self-directed cyberattacks.

These concerns have also reached Congress, where lawmakers have proposed ideas including new regulations and mechanisms to shut down AI systems, though the chances of passing broad legislation remain slim amid political divisions and the approaching midterm elections.

Independent Oversight

Amodei proposes granting independent safety evaluators access to AI companies at levels comparable to employees, to verify model safety before release.

He said Anthropic will allow independent reviewers to enter the company and examine its new systems—a concept Altman said he supports implementing at OpenAI, while Musk described peer review among AI developers as a step worth taking.

Amodei also called for a governmental role in overseeing advanced AI systems, arguing that regulation gives the public and elected officials a voice in defining the limits of what private companies can do.

As an indicator of how safety concerns are influencing corporate decisions, Altman stated that OpenAI will not go public in 2026, despite earlier speculation it might do so this year.

He explained that going public now would be "inappropriate," noting that safety issues and model alignment must be addressed first. He added he would not accept a 10% risk that the technology could cause human deaths by the end of the decade.

Meanwhile, Anthropic continues to plan an IPO, with reports indicating the company has chosen Nasdaq for its expected listing—signaling that ongoing competition among AI firms remains a key factor in its investment and business decisions.

The difficulty of achieving widespread slowdown grows as competition with Chinese companies persists, with China continuing to launch new models and narrowing the gap with American firms.

Industry officials and experts argue that any unilateral slowdown by American companies could give China an opening to advance, making calls for safety directly tied to geopolitical competition.

OpenAI and Anthropic find themselves facing a complex dilemma: strengthening safety protocols in line with increasingly powerful AI, while simultaneously maintaining development speed, investment, and the competitive edge that Washington views as central to its strategic rivalry with China.