Group 1 - The G7 summit in Alberta, Canada, released the "AI for Prosperity Declaration," focusing on the benefits and opportunities of artificial intelligence, while neglecting the term "safety" entirely [1][2] - The shift in G7's AI policy reflects a broader realignment among Western democracies, moving from early concerns about AI risks to an emphasis on its economic benefits [1][3] - The 2025 declaration significantly reduced previous global concerns about AI risks, only mentioning issues related to the power grid and the risk of being excluded from the current technological revolution [2][3] Group 2 - The trend of downplaying AI risks is not isolated to the G7 but represents a larger shift in global AI dialogues, as seen in the 2023 UK AI Safety Summit and the 2024 Seoul AI Summit [3][4] - NATO's recent policy revisions have also shifted focus from AI risks to the need to "win the technology adoption race," indicating a higher tolerance for emerging technology risks [5][6] - The U.S. Congress is considering banning most state AI laws without replacing them with federal legislation, further illustrating the diminishing emphasis on safety in AI discussions [5][6] Group 3 - The transition in AI policy is driven by multiple factors, including the change in U.S. administration and the influence of industry interests, as seen in France's shift towards self-regulation in AI [6][7] - The AI industry is experiencing a new sentiment of fear of missing out on opportunities, with significant investments flowing into AI companies like OpenAI and Mistral [8][9] - Governments are recognizing AI's potential for enhancing military capabilities, with major contracts awarded to companies like Anthropic and Google, indicating a focus on economic and military advantages [10][11] Group 4 - The lack of effective international governance structures for AI risks poses a significant challenge, as there are no mature institutions like the IAEA for nuclear technology [12][13] - The rapid development of AI may outpace the ability of nations to establish effective multilateral policies, raising concerns about emerging global AI risks, including biological threats and misinformation [12][13] - The abandonment of safety considerations in AI policy could represent a significant gamble, as the balance between caution and optimism becomes increasingly difficult to achieve [13]
从Safety到Security：西方叙事下全球AI安全治理淡化
3 6 Ke·2025-08-20 12:12