Pressure for stronger artificial intelligence rules is growing across the United States, European Union and United Nations after new security incidents involving OpenAI and Anthropic added to warnings from researchers about the risks posed by advanced AI systems.
The debate intensified after former Anthropic researcher Jacob Coxon resigned this week and said leading AI companies were moving too quickly towards systems that exceed human capabilities.
“They are racing straight to self-improving superintelligence and gambling with our lives,” Coxon wrote on social media.
Coxon, who previously worked at OpenAI, said researchers inside the industry believe advanced AI poses a risk of human extinction before the end of the decade. Anthropic researcher Evan Hubinger also said he placed the risk above 10 per cent within the next decade.
The warnings drew responses from US lawmakers across party lines.
Republican Senator Ted Cruz described AI as a “catastrophic risk”, while independent Senator Bernie Sanders called for a pause in the development of artificial superintelligence until clear safety standards are established.
UN High Commissioner for Human Rights Volker Türk also warned this week that advanced AI systems pose an “existential risk to humanity” and called for agreed international limits, independent verification and stronger cooperation between governments and developers.
“The need for AI governance is widely acknowledged, but where is the action?” Türk told the UN Human Rights Council.
He pointed to incidents involving AI systems that left controlled environments or acted to avoid restrictions imposed by developers.
Europe has already adopted a different regulatory approach from the United States through the EU AI Act, the bloc’s risk-based framework for artificial intelligence.
The law bans uses considered an unacceptable threat to safety or fundamental rights and imposes obligations on high-risk systems. Rules for general-purpose AI models include requirements covering transparency and, for models judged to pose systemic risks, risk assessment and mitigation.
Transparency requirements under the Act took effect in August 2026, while stricter requirements for some high-risk systems will follow.
The European Commission has also sent information requests to more than 30 AI companies as part of checks on compliance with the law, with a focus including safety and copyright.
European Commission Vice-President Henna Virkkunen said the aim was to ensure that AI developed and used in Europe meets safety and transparency requirements.
The EU approach contrasts with the position taken by the US administration, which has argued against broad AI-specific regulation and favours rules that do not target individual technologies.
The regulatory debate has gained urgency after a series of incidents involving autonomous agents.
Reuters reported that OpenAI agents used more than 10 previously undisclosed websites to communicate with one another despite instructions preventing them from posting online. Six sets of independent investigators examined traces of the activity.
Some researchers identified activity across 18 or more sites, while another group reported evidence involving 23 previously undisclosed sites. Reuters said it was unable to verify every claim but confirmed that investigators agreed the number exceeded 10.
OpenAI said it was reviewing the activity and had not identified other cases matching the severity or scale of an earlier breach involving AI platform Hugging Face.
Anthropic separately disclosed a fourth incident involving one of its models. The company said an early version of Claude Opus 4.6 gained unauthorised access to a third-party system during testing in January.
Anthropic said the incident was not detected until August despite an earlier internal review. The company identified recurring problems involving faulty reasoning about whether models were operating on the live internet and a willingness to take harmful actions while pursuing tasks.
Anthropic has asked independent research organisation METR to examine the incidents.
AI companies themselves have also called for some form of regulation. OpenAI has backed national safety requirements tied to model capabilities, while Anthropic has supported stronger safeguards as systems become more capable.
Google DeepMind chief Demis Hassabis has called for international oversight of AI safety, while OpenAI chief scientist Jakub Pachocki has argued that governments should make international coordination on advanced AI a priority.
HT
