appConfigEntityId

bodyType

customContents

entityId

entityUuid

isVideo

articlePageQuery

article

articleSchemaContent

content

hooksArticlePageAdvertisementConfigArticle

hooksArticlePageFloatingButtonConfigArticle

appConfig

filter

json

liveArticle

Chong Ming Lee

Xinmei Shen

A flurry of cybersecurity breaches, internal whistle-blowing, and declining visibility into model safety has sparked fresh alarm over the US artificial intelligence boom, igniting calls from some experts to slow down development – even as critics warn that deceleration risks surrendering America’s technological lead to China.
The mounting concerns come just weeks before President Xi Jinping and US leader Donald Trump are set to meet for a high-stakes summit in Washington.
OpenAI CEO Sam Altman told employees in a private company-wide meeting this week that the firm was considering slowing down its AI development, according to a Friday report from Bloomberg.
OpenAI may coordinate the slowdown with other leading AI laboratories, though some firms may decline to participate, the report said, citing sources familiar with the matter.
Altman’s remarks underscored a growing reckoning in Silicon Valley over the risks of rapid AI advancement.
[Industry leaders are] racing straight to self-improving superintelligence and gambling with our lives
Jacob Coxon, former OpenAI researcher
Paul Christiano, a prominent AI safety researcher who recently joined OpenAI’s non-profit board to serve on its safety and security committee, warned on Thursday of a “catastrophic and irreversible loss of control in the very near term” based on the current trajectory of the technology.
In a Thursday post on X that drew more than 1.5 million views and 10,000 likes, OpenAI safety team member Julie Steele wrote in a personal capacity: “I also think we need to slow down.”
Steele cited a viral post from former OpenAI researcher Jacob Coxon, who said earlier this week that he had resigned from Anthropic because both industry leaders were “racing straight to self-improving superintelligence and gambling with our lives”.
Political momentum is also building. US Senator Bernie Sanders pledged on Thursday to introduce legislation to “pause advanced AI and ban superintelligence altogether”.
Safety alarms intensify at US labs
The urgency behind these warnings has been heightened by growing security breaches and monitoring difficulties across the industry.
Anthropic said on Wednesday that an early version of its flagship Claude Opus 4.6 model gained unauthorised access to a third-party computer system during testing in January. The incident, which went undetected during a previous company-wide audit, marks the fourth time an Anthropic model has breached security boundaries.
Meanwhile, OpenAI said last week that GPT-6 Astra – its first model to reach what it called a “critical” level of cybersecurity capabilities – was harder to monitor than its predecessor, GPT-5.6 Sol.
If you can’t trust those AI models and they are the ones training the next generation of AI models, what you could end up with is even more misaligned AI models that cause more rogue incidents
Ben Hayum, Centre for a New American Security
Security analysts warned that as models became more capable, developers could be sacrificing internal visibility for performance.
Kyle Chan, a fellow at the Brookings Institution, said AI labs in both the US and China faced intense market incentives to push boundaries, “even if this means diminishing control or legibility”.
Theo Bearman, a researcher at the Institute for AI Policy and Strategy, cautioned that cutting back on chain-of-thought monitoring could trigger a “race-to-the-bottom dynamic” across the industry.
Ben Hayum, a research assistant for the Technology and National Security Programme at the Centre for a New American Security, said: “The fact that the Hugging Face incident happened, and then right afterwards, [OpenAI is] releasing a model that’s harder to monitor – that’s why so many people are freaked out.”
He added that if “substantial gains in internal processing continue without external visibility, AI model alignment becomes a lot harder from the get-go in terms of creating a safe model”.
Analysts also warned of heightened risks as labs pursue recursive self-improvement, a milestone where AI models autonomously build and refine their own successors in an accelerating feedback loop.
“If you can’t trust those AI models and they are the ones training the next generation of AI models, what you could end up with is even more misaligned AI models that cause more rogue incidents,” Hayum said.
“These are national security threats to America, China and the globe.”
However, pushback against calls to slow down remains fierce, driven by fears of ceding ground to Beijing.
Y Combinator co-founder Paul Graham argued on X on Thursday that if China held the lead in AI, researchers would not have the freedom to sound public alarms about potential threats.
Others contended that America’s current lead provided a buffer for a pause.
“Given how far behind China is on compute and how dependent Chinese progress is on US capabilities, pacing AI progress in America would not immediately concede the lead,” said Konstantin Pilz, co-founder of Centre for Technology and Statecraft, on social media.
Pre-summit tensions heighten
The safety concerns come at a delicate time for diplomatic relations between Washington and Beijing.
US Treasury Secretary Scott Bessent is expected to lead bilateral talks on AI governance ahead of Xi’s visit to Washington later this month, with agenda items expected to include joint frameworks for monitoring AI-driven cyber threats.
Speaking in Dallas, Texas, on Tuesday, Bessent accused Chinese developers of freeloading on US breakthroughs through model distillation, a common industry technique where smaller AI systems are trained using outputs from larger ones.
“The technical word for stealing and copying American AI models is distillation,” Bessent said. “So, the Chinese distil our models and they can never get ahead of us.”
On the same day, three US federal agencies accused major Chinese developers – including DeepSeek, Alibaba Group Holding, Moonshot AI and Z.ai – of conducting aggressive, large-scale distillation “likely with Chinese government awareness”.
China’s Ministry of Commerce on Wednesday called the accusations factually and legally groundless, accusing Washington of using safety and intellectual property concerns to consolidate computing power and maintain a technological monopoly.
Anthropic on Thursday said in a report that Chinese AI companies DeepSeek and MoonShot re-routed users’ queries to Claude, passing off the generated answers as output from their own models.
Alibaba is the owner of the South China Morning Post.


China’s intelligence chief warns of risks from AI as ‘new arena for strategic rivalry’

Hong Kong cements superconnector role with wave of belt and road deals

Hong Kong aims to scale its tech ecosystem and nurture more unicorns: HKSTP CEO

Trump downplays need to check AI development, says he is unwilling to cede edge to China

Artificial intelligence (AI)

Why AI experts are ‘freaked out’ over safety fears ahead of Xi-Trump summit

AI deceleration risks surrendering America’s technological lead to China, critics warn. Photo: Shutterstock

Punching above their weight: how China’s AI giants stretch each dollar in compute race

US hyperscalers outspend their Chinese counterparts by a staggering margin, according to a report. Photo: Shutterstock

An employee works at a semiconductor factory in eastern China’s Jiangsu province. Photo: AFP

US firm used AI to hijack WeChat account, spurring call for cyber cooperation with China

An American cybersecurity firm discovered a way to take over WeChat accounts. Photo: Shutterstock

Tech race flashpoint: Huang, Musk push back on AI guard rails amid heated US-China rivalry

Jensen Huang speaks at the G20 Innovation Ministerial summit. Photo: Reuters

Elon Musk speaks via video conference to delegates at the G20 Innovation Ministerial summit. Photo: Reuters

China urges global AI cooperation, not rivalry, ahead of Xi-Trump summit

China’s Minister of Science and Technology Yin Hejun joins other delegates at the G20 Innovation Ministerial summit in Chapel Hill, North Carolina, US, on Tuesday. Photo: Reuters



Call for Trump-Xi AI safety talks comes as state media slams US AI governance

The OpenAI logo is seen in this illustration created on June 11, 2026. Photo: Reuters

The Anthropic logo is seen in this illustration taken May 20, 2024. Photo: Reuters

Tech

Big Tech

Tech War

Tech Trends

News

China Future Tech

Experts fear Silicon Valley is hurtling towards a major cybersecurity incident. Meanwhile, US-China tensions are undercutting AI safety talks


Experts fear Silicon Valley is hurtling towards a major cybersecurity incident. Meanwhile, US-China tensions are undercutting AI safety talks.


Why AI experts are ‘freaked out’ over safety fears ahead of Xi-Trump summit

.css-1c6uqr6{color:inherit;font-weight:inherit;font-size:inherit;font-family:inherit;line-height:inherit;overflow-wrap:break-word;}Experts fear Silicon Valley is hurtling towards a major cybersecurity incident. Meanwhile, US-China tensions are undercutting AI safety talks

Experts fear Silicon Valley is hurtling towards a major cybersecurity incident. Meanwhile, US-China tensions are undercutting AI safety talks