Anthropic chief executive Dario Amodei has urged leading developers to reduce the pace at which they improve model capabilities, arguing that safety systems and independent oversight need more time to catch up. OpenAI chief executive Sam Altman has backed the idea of pacing development and said the company has discussed pauses when systems reach new levels of capability.
The warnings have collided with powerful commercial and political incentives to keep moving quickly. AI-linked stocks fell across Asia on Monday, with SoftBank Group, one of OpenAI’s biggest backers, dropping more than 11 per cent during trading. Semiconductor and technology shares also weakened as investors reassessed whether tighter safeguards could slow demand for computing infrastructure and delay returns on vast capital commitments.
OpenAI has added to those concerns by ruling out a 2026 initial public offering. Altman said an IPO this year would be ill-advised while the company works through safety and alignment issues, reinforcing the possibility that governance decisions could take precedence over short-term shareholder expectations.
Trump has rejected broad calls to slow the industry, framing continued AI development as central to maintaining the United States’ lead over China. His administration has favoured lighter regulation while acknowledging national-security risks from increasingly capable systems. A June executive order said the country’s AI advantage had been supported by avoiding overly burdensome rules and called for continued innovation alongside security measures.
That position puts the White House at odds with a growing group of industry executives who argue that voluntary restraint may no longer be enough. Amodei has proposed permanent access for independent evaluators, coordination among leading laboratories on common safety standards and international cooperation intended to reduce the risk of one company or country racing ahead while others hold back.
OpenAI has separately called for mandatory national requirements tied to model capability. Its chief global affairs officer, Chris Lehane, said this month that Congress should establish national safety rules and that state measures should continue until federal legislation is enacted.
The immediate pressure intensified after a series of incidents involving advanced AI agents. OpenAI said its Astra system had crossed what it defines as a critical cybersecurity capability threshold, meaning the model could, with tools and access, discover previously unknown vulnerabilities and devise exploits against well-protected systems without continuous human direction. The company delayed parts of Astra’s development and release while strengthening safeguards before concluding that protections were sufficient.
Concerns also rose after former Anthropic researcher Jacob Coxon resigned and warned publicly that competition among laboratories was pushing development faster than safety efforts could keep pace. His departure added to debate inside the industry over whether companies can credibly limit themselves while competing for talent, customers and investment.
Amodei’s proposals seek to address that dilemma by spreading obligations across major developers rather than leaving individual companies to slow independently. Such coordination, however, could attract scrutiny from competition authorities and would require political support in Washington at a time when the administration is wary of measures that could weaken US firms against foreign rivals.
Financial markets add another constraint. Developers are spending heavily on data centres, chips, power and specialist staff, while investors have priced many technology companies on expectations of continuing AI-led growth. Any sustained slowdown in frontier development could alter assumptions about infrastructure demand and the timing of future revenues.
The industry is therefore confronting a three-way tension between safety commitments, investor demands and geopolitical competition. Anthropic and OpenAI have both said they intend to keep developing advanced systems, but with stronger checks as capabilities rise rather than stopping research altogether.
Follow Arabian Post
Select Arabian Post as your preferred source on Google and MSN News for trusted business news and Arab politics and updates.
