The impact of AI on public safety The policy outlines various applications of AI that have the potential to impact public safety and human rights. It mandates that agencies implement safeguards by December 1st, including measures to mitigate algorithmic discrimination and ensure transparency in government AI usage. Agencies are required to discontinue the use of AIs that fail to meet these safeguards. Additionally, they must inform the public about any AI that is exempt from complying with the OMB policy and provide an explanation for such exemptions. Certain AIs that have a direct impact on safety, such as those controlling dams, electrical grids, traffic control systems, vehicles, and workplace robots, fall into the category of safety-impacting AIs. On the other hand, AIs that suppress or remove protected speech, generate risk assessments for law enforcement, and perform biometric identification are classified as rights-impacting.

AI-driven decisions pertaining to healthcare, housing, employment, medical diagnosis, and immigration status are also considered rights-impacting. Under the OMB policy, agencies are also encouraged to release government-owned AI code, models, and data, as long as it does not pose a risk to public safety or government operations. The new policy has received mixed reviews from various human rights and digital rights organizations. While the American Civil Liberties Union (ACLU) views the policy as an important step towards protecting US residents from AI abuses, they have also raised concerns about its limitations. The ACLU specifically highlighted broad exceptions for national security systems and intelligence agencies, as well as exemptions for sensitive law enforcement information. In a statement, Cody Venzke, Senior Policy Counsel at the ACLU, stated, "Federal uses of AI should not undermine rights and safety, but harmful and discriminatory uses of AI by national security agencies, state governments, and others remain largely unregulated. Policymakers need to take action to address these gaps and establish the necessary protections. "