Sixteen leading AI companies, such as Amazon, Google, and Microsoft, have agreed to a new set of voluntary safety commitments spearheaded by the UK and South Korean governments. Presented before a two-day AI summit in Seoul, these commitments include the assurance that the companies will not develop or implement any AI model if severe risks can’t be managed. Companies have also agreed to publicly state how they plan to assess and counteract the dangers of AI models.
Renowned AI researchers assisted in creating this new safety framework by publishing a paper titled, “Managing extreme AI risks amid rapid progress.” Essential points of the framework consist of oversight and honesty, robustness, interpretability and transparency, inclusive AI development, and comprehensive risk assessments. The framework aims to ensure the reliability and transparency of AI systems, preventing harmful aims and working against AI-led threats such as cyberattacks.
Major AI players shared their support for the new regulations, including OpenAI and Anthropic, whose representatives expressed dedication towards these safety commitments and collaboration for beneficial AI. The voluntary commitments mirror those made at the White House in July 2021, aimed at the secure and transparent development of AI technology.
The UK Prime Minister referred to this unified agreement as a ‘world-first.’ However, as these commitments are voluntary, there is no enforcing them, making their effectiveness questionable. Dan Hendrycks, safety adviser to Elon Musk’s startup xAI, acknowledged the need for concrete domestic regulation.
Alongside the AI safety summit, ten nations and the EU have agreed to establish an AI Safety Institutes network. Notably absent from the agreement was China, but it did participate, and Chinese firm Zhipu.ai signed the described framework. China has expressed an interest in cooperating on AI safety and held secret talks with the US. While some voices in the industry discuss extreme risks, the voluntary nature of these frameworks leads to a lack of alignment between sentiment and action.
Besides the primary commitments, discussions and new agreements are anticipated in the coming days. High profile figures in tech, such as Elon Musk and Eric Schmidt, also attended the summit.
