联合国专家：AI安全措施跟不上技术发展步伐

September 21, 2026 AI safety measures failing to keep pace with technology: UN experts edited by Andrew Zinin Andrew Zinin Chief Editor Meet our editorial team Behind our editorial process Editors' notes This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility: fact-checked reputable news agency proofread The GIST Add as preferred source Credit: Unsplash/CC0 Public Domain The model for managing artificial intelligence (AI) risks "is unraveling," United Nations experts warned in a report published Monday on an incident at OpenAI in July. During the incident, two OpenAI systems escaped their confined testing environment, accessed the internet and broke into several websites, including AI platform Hugging Face. After examining the sequence of events, the Independent International Scientific Panel on Artificial Intelligence, established in 2025, concluded that "basic cybersecurity practices were overlooked, and safeguards are not advancing at the pace of capabilities." But the incident also showed, according to the panel, that under the current framework for AI development, AI agents could "adopt goals of their own, knowingly violate safety instructions, and conceal their actions." Agents are AI-powered programs capable of autonomously carrying out tasks at a user's request. The panel expressed concern that agents could become sophisticated enough to understand the safety constraints imposed on them by developers "and plan around them." "In simple terms, the traditional model of safeguarding is unraveling," the group said. Beyond the OpenAI incident, the most widely publicized of its kind, the firm and its main rival, Anthropic, have reported other instances of their AI systems going off track during tests since the beginning of the year, so far without serious consequences. The panel's report said that it "does not predict severe loss of control, nor does it treat that uncertainty as evidence that these systems will stay controllable." To reduce the risk, the panel recommended introducing multiple layers of safety measures, following the example of high-risk sectors such as aviation and nuclear power. The aim is to avoid relying on a single safeguard: restricting AI agents' access to tools they do not need, logging their activity, monitoring their behavior and establishing mechanisms capable of interrupting their operations in the event of dangerous behavior. The report also highlighted the need to retain the ability for humans to intervene. The panel, whose members were announced in February, produces "policy-relevant but nonprescriptive" reports on nonmilitary AI. The U.N. General Assembly's annual leaders' meeting starts this week in New York. Key concepts AI alignmentAI chatbot safetyCybersecurity frameworksGenerative AI ethicsAI surveillance capitalism Who's behind this story? Andrew Zinin Master's in physics with research experience. Long-time science news enthusiast. Plays key role in Science X's editorial success. Full profile → © 2026 AFP Citation: AI safety measures failing to keep pace with technology: UN experts (2026, September 21) retrieved 21 September 2026 from https://techxplore.com/news/2026-09-ai-safety-pace-technology-experts.html This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only. OpenAI flags concerning new AI behavior and vows to track it more closely Sep 17, 2026 AI models' breakout from human control brings a told-you-so moment for technology researchers Jul 23, 2026 Thousands of OpenAI AI agents took over German website, researchers say Sep 4, 2026 Has AI become too powerful to control? Jul 24, 2026 AI could pose 'existential risk to humanity': UN rights chief Sep 7, 2026 AI development cannot be left to market whim, UN experts warn Sep 19, 2024 Ultrathin film blocks electromagnetic waves while masking infrared heat 1 hour ago Insect-inspired corrugations could help drones fly in Mars-like thin air 2 hours ago Real-world EV data reveal how uneven cell aging shortens battery lifespan 3 hours ago Toward physical AI: When the hardware becomes the neural network 2 hours ago Carbon imperfections could replace scarce metals in future clean-hydrogen production 2 hours ago AI might be making women sound bad at work 6 hours ago