White House demands ‘concrete safeguards’ from federal agencies while using AI

US govt on AI: The White House announced that federal agencies using artificial intelligence (AI) must implement "concrete safeguards" by December 1 to protect the rights of Americans and ensure safety as the government expands AI applications across various sectors.

The Office of Management and Budget issued a directive instructing federal agencies to monitor, evaluate, and test AI's effects on the public, mitigate algorithmic discrimination risks and improve transparency regarding government AI usage. Agencies are also mandated to conduct risk assessments and establish operational and governance metrics.

Safeguards to ensure privacy

To ensure accountability, agencies must implement specific safeguards when using AI that could impact the rights or safety of Americans, including comprehensive public disclosures outlining the government's AI usage.

US President Joe Biden previously signed an executive order in October leveraging the Defense Production Act to compel developers of AI systems with potential risks to national security, the economy, public health, or safety to share safety test results with the US government before public release.

New safeguards will enable air travellers to opt out of Transportation Security Administration facial recognition screening promptly. Additionally, federal healthcare AI applications must be overseen by humans to verify diagnostic decisions.

Usage of Generative AI to get public

The use of Generative AI, which can generate text, images, and videos based on open-ended prompts, has elicited both excitement and concerns regarding potential job displacement, electoral interference, and adverse effects on society.

Government agencies are now required to publish inventories of AI use cases, report AI usage metrics, and release government-owned AI code, models, and data that pose no risks.

The Biden administration highlighted ongoing federal AI initiatives, such as FEMA using AI to assess structural hurricane damage, the CDC employing AI to predict disease spread and detect opioid use, and the FAA using AI to optimise air traffic flow in major cities.

Furthermore, the White House plans to recruit 100 AI professionals to promote safe AI usage and mandates federal agencies to appoint chief AI officers within 60 days.

In January, the Biden administration proposed requiring U.S. cloud companies to implement "know your customer" rules to ascertain whether foreign entities are accessing U.S. data centres for AI model training purposes.

(With Reuters inputs)