AI Watch: 🚨 Shadows & 30 Days 🚀

AI MODEL REVIEW FRAMEWORK: A NEW APPROACH
The Biden administration has swiftly responded to concerns surrounding the rapid development of artificial intelligence by enacting an executive order aimed at establishing a robust framework for evaluating AI models. This initiative, spearheaded by the Office of the National Cyber Director, prioritizes proactive risk mitigation and information sharing, moving away from the previous administration’s largely ideological approach to AI regulation. The core of the order centers on enabling the federal government to collaborate with critical infrastructure operators – encompassing sectors such as banking, utilities, and healthcare – to identify and address vulnerabilities within AI systems before they are publicly released. This represents a significant shift in strategy, recognizing the potential security implications of advanced AI technologies.

INFORMATION SHARING AND VULNERABILITY MITIGATION
A key component of the new framework involves a voluntary review process for AI companies. The executive order mandates that companies with powerful models will be invited to share their technology with the government for a 30-day assessment. This contrasts sharply with an earlier draft that proposed a 90-day review period, reflecting a pragmatic adjustment based on industry feedback. The aim is to allow the government to proactively identify and address potential vulnerabilities within these models, enabling operators of critical infrastructure to implement necessary patches and safeguards before widespread public access. The Center for Democracy and Technology (CDT) views this approach favorably, emphasizing the importance of testing, particularly for critical infrastructure, to identify and rectify vulnerabilities. Samir Jain, the organization’s vice-president of policy, highlighted the sense in this approach, noting the potential for improved security and the ability to “patch them before the capabilities become widely available.” However, Jain also expressed concerns about the order's opacity, noting the lack of public visibility into the benchmarking process.

SHIFTING ADMINISTRATION POLICY AND SECURITY PRIORITIES
The decision by President Biden to regulate the AI industry marks a notable departure from the previous administration’s stance. The earlier AI Action Plan from the Trump administration focused on ideological concerns, specifically targeting AI systems deemed “woke” and those promoting “ideological dogmas” like DEI. Furthermore, the administration actively sought to prevent states from enacting their own AI restrictions, even to the point of establishing a DOJ task force to challenge these laws. Following a realization of the genuine security risks posed by AI, the administration has shifted its focus to proactively mitigating those risks. As Samir Jain stated, “In that sense, the executive order is a change from the perspective the administration has realized that AI poses real security risks and the government needs to act to mitigate or address those risks.” This represents a fundamental change in priorities, moving away from a purely ideological stance towards a more security-conscious approach to AI regulation.