Industry groups are challenging the Treasury Department’s use of definitions from President Biden’s Executive Order on Artificial Intelligence (AI) to set restrictions on outbound investments aimed at limiting China’s access to advanced technologies. The Semiconductor Industry Alliance and other business groups argue that applying the broad AI definitions from the October 2023 Executive Order on AI to the August 2023 Executive Order on Outbound Investments could result in overly expansive and imprecise regulations. These groups contend that the definitions, originally intended to promote safe and secure AI, may inadvertently capture entities that merely integrate AI without developing the underlying technology, potentially leading to outdated policies in the rapidly evolving AI landscape.

The dispute highlights potential tensions between Biden’s 2 Executive Orders, both designed to maintain U.S. leadership in critical technologies. Industry representatives are urging the Treasury to narrow the scope of its proposed rule to focus more precisely on AI applications that pose genuine national security risks, rather than broadly applying definitions that could affect a wide range of products and technologies. They argue that this approach would align better with the Administration’s broader AI policy goals and avoid unnecessary disruptions in the technology sector.

The RAND Corporation has released 2 reports that explore the European Union’s AI regulatory framework and offer recommendations for U.S. policymakers to better align trans-Atlantic AI regulations. The studies emphasize the importance of collaboration between the EU and U.S., given the global impact of regulatory developments in these influential regions. The reports focus on lessons from the EU’s AI Act, which provides a comprehensive regulatory framework for AI, and suggest how the U.S. might adopt similar standards, particularly in the areas of general-purpose AI models, risk management, and privacy protections.

One report specifically examines how the EU and U.S. differ in their classification and regulation of AI models with systemic risks, offering options for the U.S. to align more closely with EU practices in standards-setting and reporting. The 2^nd report addresses privacy concerns related to AI, highlighting the differences between U.S. and EU approaches to data protection. RAND suggests that the U.S. could benefit from adopting measures such as data minimization, mandatory audits, and transparency requirements to ensure that AI innovation proceeds responsibly without compromising privacy.

The Senate Appropriations Committee, through its recently approved $851.7 billion defense spending bill, is advocating for the increased use of AI to streamline the Department of Defense’s operations and cut costs. The Committee highlighted that the Pentagon’s reliance on outdated systems and practices is inefficient, wasting resources and hindering recruitment and retention efforts. The report accompanying the bill stresses the need for the Department of Defense (DOD) to adopt AI and other modern technologies, which are already widely used in the private sector, to improve its business operations and reduce its civilian workforce costs by over $1 billion.

The Committee’s report also calls for the development of a “hedge portfolio” that utilizes AI and advanced technologies to address tactical and logistical risks, recommending an initial $1 billion investment. Additionally, the report mandates that the Secretary of Defense deliver a plan within 120 days outlining specific goals and metrics for the adoption of AI and other technologies across the department, signaling a strong push for modernization at the Pentagon.

At the recent Black Hat and Def Con Conferences, cybersecurity leaders strongly advocated for holding software makers liable for vulnerabilities in their products unless they adhere to cybersecurity best practices. Cybersecurity and Infrastructure Security Agency (CISA) Director Jen Easterly emphasized that software flaws should be considered “product defects” and argued that market forces alone are insufficient to ensure software safety. She reiterated her call for legal liability for software companies, with safe harbors for those following secure-by-design principles. National Cyber Director Harry Coker echoed this sentiment, signaling that a software liability proposal is forthcoming. The Biden Administration’s national cyber strategy also highlights the need for such legislation, aiming to establish higher standards of care and prevent software makers from disclaiming liability by contract.

AI’s role in cybersecurity was another focal point, with leaders acknowledging that while AI tools, like large language models, can introduce vulnerabilities, they are also crucial for enhancing security. Veracode CTO Chris Wysopal noted that AI-generated code is as prone to security flaws as human-written code, but other AI tools will be key in addressing these issues. The discussions underscored the pressing need for both stronger software accountability and the strategic use of AI to bolster cybersecurity efforts.

The Consumer Financial Protection Bureau (CFPB) has strongly opposed the idea of “regulatory sandboxes” for artificial intelligence, where foundational laws are temporarily suspended to encourage innovation. In comments submitted to the Treasury Department, the CFPB argued that innovation should not come at the expense of consumer protections, pointing to past experiments with such sandboxes under the Trump Administration that fell short of their intended goals. The CFPB emphasized that all market participants should adhere to the same set of rules, ensuring a level playing field and protecting consumers from potential discrimination. The agency also highlighted the importance of maintaining transparency and explainability in AI-driven decision-making processes, particularly in sectors like lending, where AI models are increasingly used.

Microsoft and Palantir are collaborating to offer advanced AI services for classified government use, with both companies optimistic about securing final approval from national security and intelligence agencies. The partnership aims to meet the Biden Administration’s stringent AI procurement standards, as outlined in the October 2023 Executive Order on AI, which emphasizes safety and rights protection. While no contracts have been signed yet, Microsoft’s OpenAI-enabled Azure Cloud platform has been approved for unclassified government use, and the companies expect to receive clearance for classified use within the next few months. The collaboration will allow the U.S. Defense and Intelligence Community to utilize AI-driven operational workloads in secure environments, marking a significant advancement in AI capabilities for national security missions.

Pressure is intensifying on California state Sen. Scott Wiener (D-CA) to either scale back or abandon his groundbreaking AI regulatory bill, SB 1047, as opposition mounts from key lawmakers, industry leaders, and tech companies. U.S. Representative Zoe Lofgren (D-CA), a prominent figure in technology policy, expressed serious concerns about the bill, arguing that it imposes regulations without sufficient evidence and could harm California’s innovation economy. The bill, which aims to establish strict safety and liability requirements for AI developers, has sparked significant criticism, particularly from tech giants like Meta, which argues that the bill misunderstands the AI development process and could stifle innovation in the state.

The bill is facing increasing resistance from the tech industry, including influential figures who warn of unintended consequences and a potential exodus of AI development from California. Industry groups argue that SB 1047’s provisions, such as holding developers liable for downstream harms and requiring 3^rd-party audits, are overly burdensome and could discourage AI innovation. The fiscal impact of the bill, estimated to cost the state tens of millions of dollars annually, adds to the controversy, with Governor Gavin Newsom’s (D-CA) potential veto looming due to budgetary concerns. The bill is currently awaiting a vote in the Assembly Appropriations Committee, with a deadline of August 16^th to clear fiscal hurdles.

Clients, friends and family only.