Anthropic CEO Urges Slower AI Development for Safety Under UN Supervision

Dario Amodei, CEO of Anthropic, stated clearly in an address today at a United Nations Security Council (UNSC) meeting that the company will slow down the research, development, and deployment of artificial intelligence (AI) technology when necessary to ensure the safety and controllability of next-generation models. Amodei pointed out that the AI industry is in a phase of rapid growth, with the capabilities of everything from language models to generative image systems continuously improving, while simultaneously bringing risks such as the spread of misinformation, deepfakes, and decision-making bias. As one of the few companies in the industry with safety as its core mission, Anthropic is willing to strike a balance between technological breakthroughs and risk management, and is open to accepting supervision and recommendations from international organizations.

Over the past few years, a series of global AI-related security incidents have accelerated the international community's focus on artificial intelligence governance. In 2023, the United States and the European Union successively released AI risk management frameworks, requiring large-scale models to undergo transparency and testability reviews prior to public release; that same year, OpenAI's ChatGPT was warned by multiple governments for generating inaccurate information. These cases highlight that if technology developers fail to simultaneously establish risk assessment mechanisms, they may pose shocks to public safety, the information order, and even international security. In recent years, the UNSC has also begun adding AI to its agenda, taking the view that if advanced automated systems are militarized or misused, they will alter the nature of warfare and strategic balance.

Since its establishment in 2021, Anthropic has consistently made "alignment" and "safety" the core of its research and development, continuously optimizing the conversation quality and safety protections of its launched Claude series of models. The company maintains a dedicated safety team responsible for risk assessment, adversarial testing, and research into the interpretability of model behavior. Amodei mentioned in his remarks that if potential harms or unresolvable biases are discovered during future model iterations, Anthropic will proactively suspend the release of new versions, or even revert to previous, more conservative versions, to avoid unforeseen negative impacts on users or society. This move is not only a commitment to its own responsibilities, but also echoes the United Nations' calls in the field of AI ethics and safety.

UNSC meeting records show that attending representatives generally focused on the potential threats of AI in information warfare, the dissemination of false information, and the automated weaponization of technology, calling on governments and enterprises to establish cross-border cooperation mechanisms. Amodei's statement is seen as a positive response by the private sector to international governance and sets a "safety-first" example for other AI companies. Representatives from some member states pointed out that if enterprises incorporate risk assessments at the early stages of research and development and actively slow down when necessary, it will help reduce future international disputes or humanitarian crises caused by out-of-control technology.

Finally, Amodei emphasized that the long-term development of AI is inseparable from multi-party collaboration, including the joint participation of governments, academia, civil society groups, and the industrial sector. He called on the United Nations to continue promoting a global AI governance framework and to provide transparent oversight mechanisms so that enterprises do not neglect the bottom line of safety in their pursuit of innovation. This commitment is not only Anthropic's self-discipline regarding its own technological path, but also a contribution to the global AI safety ecosystem, demonstrating that amidst a rapidly evolving technological wave, a responsible research and development attitude remains key to maintaining international peace and public welfare. (Source: Central News Agency)