The news of September 22ndUnited NationsA team of scientific experts issued a warning:Before the risks are fully identified, we should be more and more capable AI AgentControlled. This oneThe report is also the first important evaluation by the international organization of the OpenAI invasion of Hugging Face earlier this year。
1AI notes that this week ' s leaders gathered in New York for the United Nations General Assembly and that both China and the United States are engaged in a parallel process of artificial intelligence-related talks; the release of this report has further put artificial intelligence issues on the global diplomatic agenda. Last week, the Secretary-General of the United Nations, António Guterres, called on Governments to work together to address the threat posed by artificial intelligence. At the same time, he warned: “The world cannot afford a vicious race to lower standards at the level of AI security.”
The paper was the first thematic briefing by an independent international scientific expert panel of the United Nations. The Panel was established last year as the United Nations “first global scientific body in the field of personal intelligence”. The briefing suggested that, even if countries chose different legislative and regulatory paths, there was a need for more global attention and resources to manage the types of risks that advanced artificial intelligence was emerging, as well as for enhanced international collaboration on security and accountability。
Of particular importance in the report is the Panel ' s suggestion that the world need not wait until the scientific community has fully clarified the mechanisms and causes of such accidents before putting in place more protective measures. In the Panel's view, the risk of loss of control is precisely the problem to be addressed by the precautionary principle: “i.e. potential damage may be catastrophic and irreversible, but the probability of the risk occurring is not yet scientifically certain”
The precautionary principle was first enshrined in the 1992 Rio Declaration on Environment and Development. The principle is clear:Uncertainty at the scientific level could not be invoked to delay the adoption of regulatory measures against the risk of serious or irreversible harmI don't know. This principle has since profoundly influenced policy development in the field of environment and public health, particularly in the EU region。
Since Hugging Face was first reported, a number of similar incidents have been documented, involving businesses such as Openai, Anthropic, Google and Meta; they include both attacks against real-world targets, as well as a plethora of smart group message boards。