At a glance.
- Artificial intelligence regulation in the EU and the US.
- US Supreme Court limits possible Federal Trade Commission enforcement actions.
- US State Department closer to a cyber diplomacy office.
AI regulation in the EU and US.
Cooley unpacks the European Commission’s sweeping new AI bill, the first of its kind in the world, which touches on facial recognition, surveillance, policing, commercial applications, product safety, and data security. The proposal will likely affect the development, promotion, and implementation by any global suppliers, distributors, or users of technology falling under “broadly defined AI approaches,” such as systems executing safety tasks with an autonomous component or customer-facing AI. The law would set standards for human oversight, data protection, and user awareness, for example when the service is predicting intentions, behaviors, or personal attributes. An “ethical by design” guideline would require baking “respect for ethical principles and human rights in the design of the AI.”
The draft law’s risk-based approach manifests in a “pyramid of requirements” relative to a technology’s risk category: unacceptable, high, limited, or minimal. Unacceptable uses include subliminal messaging, exploitation of protected groups, social scoring, and most instances of “live remote biometric identification” in public by police. According to the Wall Street Journal, uses pertaining to “critical infrastructure, college admissions and loan applications” fall into the high risk category. In general, another Wall Street Journal article says, companies would need to report information about the function and accuracy of their facial-recognition tools and accompanying reference sets.
The bill would create a European Artificial Intelligence Board and new member state regulators, which could impose fines of up to six percent of company revenue. The Wall Street Journal remarks on the high cost of compliance, including the cost of legal advice and compliance technology. Cooley notes that “there is a strong risk of stifling good innovation without sufficient benefit,” but “amendments are likely” as the bill winds through the legislature. The result is likely to chart the course for global AI regulation.
MIT Technology Review is more excited about a new memo out of the US Federal Trade Commission (FTC) that takes on algorithm bias. The blog clarifies that the FTC Act’s provisions against “unfair or deceptive practices” cover “the sale or use of – for example – racially biased algorithms.” Although the FTC can’t touch Government abuses, it can tackle businesses claiming to offer unbiased “facial recognition systems, predictive policing algorithms, or healthcare tools.” Technology Review sees the memo as indicative of a revolution in US regulatory action, and the nebulousness of the EU measure as inviting years of legal battles by design.
US Supreme Court clips the FTC's wings.
Speaking of the Federal Trade Commission, the US Supreme Court yesterday voted unanimously that the law the FTC has used for some time doesn't, as written, give the Commission authority to recover money from companies for the purpose of compensating consumers whom the FTC finds have been defrauded, the Wall Street Journal reports.
The law under which the FTC has brought such suits is known as Section 13(b). “13(b) as currently written does not grant the Commission authority to obtain equitable monetary relief,” Associate Justice Stephen G. Breyer (a Clinton appointee) wrote for the Court. He also pointed out that the FTC could seek restitution under other provisions of the law. “If the Commission believes that authority too cumbersome or otherwise inadequate, it is, of course, free to ask Congress to grant it further remedial authority.”
The FTC has responded with dismay and a manifest sense of injured merit. POLITICO's headline fairly sums up the FTC's reaction: "'The Supreme Court ruled in favor of scam artists,' FTC chief says after justices gut agency's powers." But a 9-0 decision would seem to suggest that the issue wasn't, for SCOTUS, a contentious one. The nine sitting Justices were appointed by Presidents George H.W. Bush, Bill Clinton, George W. Bush, Barack Obama, and Donald Trump, and it seems unlikely that the decision was an ideological hit job.
US House approves new State Department cyber bureau.
A bill that would set up in the State Department a “cyber diplomacy office” tasked with furthering cyber norms has cleared the House of Representatives, CyberScoop reports. If signed into law, the Cyber Diplomacy Act would establish a Bureau of International Cyberspace Policy led by a cyber ambassador.