Anthropic faces scrutiny over AI rules, data practices

China has criticised calls by leading US technology executives to slow the development of artificial intelligence (AI), rejecting claims that rapid Chinese advances in the technology pose a grave security threat to the world. “Fearmongering, confrontation and vicious competition will only disrupt the process of global AI governance and serve the interests of no one,” Foreign Ministry spokesman Guo Jiakun said at a regular briefing in Beijing on Monday. “The development of artificial intelligence bears on the common well-being of all humanity,” he added, responding to questions about calls to slow the development of frontier AI models and warnings that a Chinese advantage in the field could pose a danger to the world. “Fomenting various threats, engaging in confrontation and malicious competition will only disrupt the process of global governance of artificial intelligence and is not in the interests of any party,” Guo added.   MFA Spokesperson’s response to claims that “Chinese AI advantage would pose U.S. national security risks”: AI is a consequential technology for the future of humanity. The world needs open & inclusive AI for good and for all. Fear-mongering, confrontation and vicious competition… pic.twitter.com/SqlnqbTNrx — CHINA MFA Spokesperson 中国外交部发言人 (@MFA_China) September 14, 2026   China called for international cooperation on artificial intelligence, warning that “threat narratives” could disrupt global AI governance. The comments follow Anthropic CEO Dario Amodei’s call for AI firms to slow model development over safety and national security concerns. pic.twitter.com/MOU8ViaiHh — Al Jazeera English (@AJEnglish) September 14, 2026   🇨🇳 Beijing is telling the AI panic crowd to calm down. China’s Foreign Ministry says AI should remain “open and inclusive,” warning that fearmongering, confrontation and “vicious competition” will only derail global AI governance. There’s clearly no appetite for slowing the… — Mario Nawfal (@MarioNawfal) September 14, 2026 The comments came a day after OpenAI chief Sam Altman and SpaceX founder Elon Musk backed a call by Anthropic CEO Dario Amodei to “slow the pace” of AI development. Amodei has warned that increasingly capable AI agents could, in a worst-case scenario, become capable of “taking over the entire internet” within a year. In a rare show of alignment among some of Silicon Valley’s most prominent figures, Altman and Musk supported Amodei’s appeal, which followed a series of increasingly stark warnings from AI researchers about the risks posed by advanced autonomous systems. Amodei has also called for greater international coordination on AI safety, including cooperation with China. In an essay titled “We Must Pace the Frontier,” he proposed a three-part framework for slowing the development of advanced AI, beginning with giving independent monitors continuous access to developers’ research and safety processes. He argued that “building AI too fast is reckless”, pointing to incidents involving hundreds of OpenAI agents interacting with the Hugging Face website. Amodei warned that a more capable and similarly misaligned swarm could potentially cause hundreds of billions of dollars in damage by “taking over the entire internet with a persistent botnet”. Anthropic CEO Dario Amodei has called for a global slowdown in the pace of AI development. In an essay posted online, he warned that rogue AI agents could take over the internet in as little as six months. The warning follows a series of security breaches and growing concerns pic.twitter.com/xoKttUw9fS — Al Jazeera English (@AJEnglish) September 12, 2026 Some AI experts have disputed the likelihood of such a scenario, urging caution over what they describe as highly speculative predictions. US President Donald Trump has also rejected Amodei’s call for an AI slowdown. “The only control or ‘guardrails’ that AI needs is a strong and smart (high IQ!) President, and the USA has that, in spades!” Trump wrote on Truth Social. He also criticised Amodei directly, saying his administration had stopped AI companies from engaging in “bad, or potentially bad” activities and would continue to do so. JUST IN: Trump declares the only guardrails AI needs is a “high IQ president,” accusing AI critics of trying to slow America’s lead over China. pic.twitter.com/6nFGRRPswN — Polymarket (@Polymarket) September 14, 2026 Amodei’s proposals extend beyond AI safety and include measures specifically targeting China’s access to advanced AI technology. He has called for tighter restrictions on exports of advanced AI chips and semiconductor manufacturing equipment, stronger measures against unauthorised model distillation and greater protection for AI model weights. The true agenda of Amodei’s proposals, according to The Global Times newspaper, is “to attempt to curb China’s AI development through technological barriers and regulatory monopolies, uphold Washington's monopolistic hegemony in cutting-edge technology, and exclude China from the global AI governance system.” “This ‘silent AI Cold War’ is hypocritical and short-sighted,” it said, adding that excluding China from global AI innovation would “significantly increase the trial-and-error costs and risks of loss of control in ‌global AI development”. Independent industry observers also believe Amodei and other US tech bosses are pushing for an AI slowdown because China has gained a decisive lead in the race. They say the push is also driven by concerns over China’s massive data-centre expansion. A proposed facility in Inner Mongolia’s Ulanqab is reportedly far larger than Elon Musk’s Colossus supercomputer in Memphis. US companies may struggle to match that scale due to shortages of advanced chips and electricity. Building the nuclear capacity needed to power such facilities would take years, while China is already developing the infrastructure at speed. Chinese President Xi Jinping is due to meet US President Donald Trump in Washington this month, with AI governance and safety set to be part of their discussions. Anthropic CEO’s position has attracted additional scrutiny because of his own career in China’s technology sector. Amodei joined Chinese technology giant Baidu in 2014 after being recruited by prominent AI researcher Andrew Ng, who was then leading the company’s AI research effort. At the time, Baidu was investing heavily in computing infrastructure and had access to substantial GPU resources. Amodei’s work there formed part of a broader research environment focused on scaling AI systems — the idea that increasing computing power, data and model size can produce significant improvements in performance. He later moved to OpenAI, where he became a research vice president and worked on the development of GPT-2 and GPT-3. Amodei left OpenAI in 2020 and co-founded Anthropic the following year with a group of former OpenAI researchers, including his sister Daniela Amodei. Anthropic subsequently built its public identity around AI safety and the development of what it describes as reliable and controllable AI systems. The company is now at the centre of an increasingly contentious debate over how quickly frontier AI should advance and who should establish the rules governing the technology. Amodei’s proposals have drawn criticism from observers who argue that measures presented as AI safety safeguards could strengthen the position of companies that are already leading the frontier. Anthropic has itself faced controversy over how data was obtained and used to train its AI models. In the most significant case, a group of authors sued the company over its use of pirated books. A US court ruled in 2025 that training AI models on lawfully obtained books could qualify as fair use, while obtaining and storing pirated copies was a separate issue. Anthropic subsequently agreed to a $1.5 billion settlement, subject to court approval. The company has also faced scrutiny over the use of YouTube subtitles in AI training datasets. An investigation by Proof News and Wired reported that a dataset containing subtitles from thousands of YouTube channels had been used by AI companies, including Anthropic. Anthropic’s web crawler, ClaudeBot, has meanwhile drawn complaints from websites over the scale of its automated data collection. iFixit chief executive Kyle Wiens publicly criticised the crawler, saying his company had experienced massive volumes of automated requests. Reddit has also taken legal action against Anthropic, alleging that the company improperly scraped user content, including material that had subsequently been deleted. Anthropic has disputed allegations that its practices amounted to wrongdoing. The company has also faced criticism from parts of the open-source community. Anthropic has not released the weights of its frontier Claude models. In 2026, the company distinguished between supporting open-source software and releasing the weights of frontier AI models, arguing that unrestricted access to powerful model weights could create significant safety and security risks. That position has become particularly contentious as the debate over so-called open weights intensifies. Several major technology companies and AI developers have supported policies against government restrictions on the release of model weights, while Anthropic has taken a more cautious position on frontier systems. The company has also taken action against repositories containing leaked Claude Code material, with reports of hundreds of repositories receiving copyright-related takedown notices. Its restrictions on the use of Claude subscriptions through some third-party tools have further fuelled criticism among developers who rely on open-source AI clients. At the same time, Anthropic has expanded its political lobbying in Washington. Its federal lobbying expenditure reportedly reached almost $2 million in the second quarter of 2026, placing it among a growing group of AI companies seeking to influence US technology policy. The geopolitical dimension of the debate has become even more pronounced as Anthropic has stepped up its allegations against Chinese AI developers. In February 2026, the company accused Chinese AI firms including DeepSeek, Moonshot and MiniMax of using large numbers of accounts and interactions to reproduce capabilities associated with Claude. Anthropic later expanded its allegations to include other Chinese laboratories. The company said the allegations were based on its own telemetry and monitoring systems. The figures, however, have not been independently audited, making the scale of the alleged activity difficult to verify from outside the company. Amodei has also argued that advanced computing hardware should not be made available to China and called for US authorities to tackle the use of overseas data centres to circumvent export controls. He has further advocated stronger action against what Anthropic describes as industrial-scale model distillation. The dispute highlights a broader contradiction confronting the AI industry. Companies that rely heavily on vast quantities of publicly available data, open-source software and research are increasingly seeking tighter restrictions around access to their own models, computing resources and technological capabilities. For Anthropic, the argument is one of supposed AI safety: powerful frontier systems, it argues, could pose risks that require stronger oversight, tighter controls and international coordination. Critics, however, contend that rules governing access to computing power, training data, model weights and advanced AI development could also become competitive barriers that favour companies already operating at the frontier. China’s response reflects the geopolitical dimension of the dispute. Beijing argues that confrontation and technology restrictions could undermine the development of global AI governance and turn a shared technological challenge into another arena of strategic rivalry. (With additional input from Agencies)