According to BlockBeats, on July 17th, the semiconductor sector came under pressure as investors reassessed AI deals. Dark Side of the Moon's claim that its Kimi K3 model could compete with models from OpenAI and Anthropic sparked renewed concerns about AI company valuations and chip spending prospects.

Despite the sell-off in chip stocks, the overall market breadth remains healthy, and the trend is more likely to reflect a rotation of funds from the semiconductor sector rather than a full-blown market retreat.

According to Beating's monitoring, Xue Lan, Dean of the Institute for International Governance of Artificial Intelligence at Tsinghua University, Nicholas B. Dirks, President and CEO of the New York Academy of Sciences, and Mark Nitzberg, Executive Director of the Center for Human-Compatible Artificial Intelligence at the Artificial Intelligence Research Lab at UC Berkeley, participated in a roundtable discussion at the 2026 World Artificial Intelligence Conference, focusing on the governance of AI agents. The panelists pointed out that AI is moving from "assisting in judgment" to "acting autonomously," becoming intelligent agents that replace humans in performing tasks—a new manifestation of the traditional principal-agent problem in the digital age. Unlike human agents, AI faces the challenge of misaligned goals and, as a black box, cannot bear legal responsibility. Accountability must shift to the entire execution chain, including developers, deployers, and regulators.

The guests reached a strong consensus: decisions involving life and death consequences, scenarios where errors are irreparable, and all issues involving ethical and value judgments must never be left to AI. Humans can authorize AI to act, but not to be responsible for it. Every authorization should be revoked, every action should be accountable, and the speed at which delegation of authority is transferred should never exceed the speed at which humans can verify AI's capabilities.

At the security mechanism level, the guests proposed that building trustworthy AI requires meeting three major engineering characteristics: solid foundation, transparent operation, and controllable use. At the institutional level, it is necessary to establish globally unified AI safety assessment standards, mutually recognized testing systems, and incident data sharing mechanisms; clearly define the red lines for AI development; establish early warning and monitoring mechanisms; and promote AI safety as a global public good. The boundaries of AI governance are not the end point of technology, but rather the starting point for humanity to rediscover its own value, responsibilities, and the direction of civilization.

---------------------------------

Click the original link below to join the Beating · Lark AI news channel and monitor global AI hot topics and news 24/7.