一份2026年全球AI安全报告警告AI快速进步的速度超过预期, 引起人们对可靠性、滥用和工作影响的关切, A 2026 global AI safety report warns rapid AI progress outpaces expectations, raising concerns over reliability, misuse, and job impacts, despite U.S. non-endorsement.

由30个国家和100名专家支持的《2026年国际AI安全报告》警告说,AI的进展快于预期,新的模型在数学和编码等复杂任务中表现出色,但行为仍然前后不一、不可靠。 The 2026 International AI Safety Report, backed by 30 nations and 100 experts, warns that AI is advancing faster than anticipated, with new models excelling in complex tasks like math and coding but still showing inconsistent, unreliable behavior. 虽然人工智能在软件工程方面正在迅速提高,并产生越来越现实的深层假冒,但人们越来越担心在网络攻击,生物武器和操纵中滥用. While AI is rapidly improving in software engineering and generating increasingly realistic deepfakes, concerns grow over misuse in cyberattacks, bioweapons, and manipulation. 尽管报告仍是独立的,但美国拒绝认可该报告。 The U.S. declined to endorse the report, though it remains independent. 调查结果强调人们对人工智能伴侣的情感依赖度日益增加, 潜在的就业乱, 敦促在德里人工智能影响峰会之前全球合作. The findings highlight rising emotional dependence on AI companions and potential job disruptions, urging global cooperation ahead of the AI Impact Summit in Delhi.