Everything Brokers Need to Know About the World's First AI Act

On May 21, 2024, the European Council gave the thumbs up to the EU AI Act after a four-year legislative journey. Typically, once the European Parliament and Council President sign off, the Act will hit the EU Official Journal and kick into gear within 20 days. This Act is a game-changer, being the world's first official regulation on AI tech globally.

A Four-Year Legislative Journey

The roots of the Act stretch back to 2020 when the European Commission issued the White Paper on Artificial Intelligence in February that year. In October of the same year, EU leaders deliberated on AI topics for the first time at the Council.

Fast forward to April 21, 2021, when the European Commission formally pitched the Act. By December 6, 2022, the European Council gave the nod to the general direction, paving the way for talks with the European Parliament. After a three-day marathon of talks from December 6-9, 2023, the European Council and Parliament finally shook hands on an agreement.

March 13, 2024, saw the European Parliament casting votes on the Act, scoring a landslide win with 523 in favor, 46 against, and 49 abstentions. Last week, as mentioned, on May 21, the Act received approval from the European Council. However, as noted, this approval doesn't signify immediate enforcement; nevertheless, it's only a matter of time.

The Act's Core Concept Is Built on Risk Management

Fundamentally, the EU Act is grounded in the concept of risk management. It doesn't aim to directly regulate AI tech itself but rather the risks tied to its use in specific societal domains. The Act categorizes risks into five levels: "Unacceptable risk," "High risk," "General-purpose AI," "Limited risk," and "Minimal risk."

Given that the Act frames AI tech as "human-centric" and shall be geared toward enhancing human well-being, certain applications are off-limits unless otherwise waived. This includes some uses that may crop up in processes like KYC by financial institutions, including brokers. For example, under "Unacceptable risk," it explicitly lists applications such as those that manipulate human behavior, real-time remote biometric identification (including facial recognition) in public spaces, and social scoring based on personal characteristics, socio-economic status, or behavior. These applications need waivers before use, impacting crucial AI applications in financial decision-making processes like credit scoring, risk assessment, and fraud detection.

Furthermore, the "High-risk" category encompasses biometric systems, the operation of critical infrastructure, and AI-supported HR software (e.g., job application software).

How Brokers Should Respond to the Act

The Act mandates comprehensive risk assessment and mitigation measures to ensure transparency, accuracy, and fairness in AI usage scenarios. For financial institutions, including brokers, to dodge non-compliance post the Act's rollout, thorough risk assessments are a must. Effective risk mitigation measures should be put in place for various AI usage scenarios in trading.

This entails ensuring data quality of AI systems, keeping detailed documentation for transparency, fulfilling disclosure obligations to investors, and setting up an effective system of human oversight.

Moreover, according to provisions regarding "Limited risk," brokers must tag audio, video, and image content generated by AI applications like chatbots and biometric classification systems as "AI-generated." This helps users decide whether to further engage, thus relieving brokers of liability.

Final Thoughts

For brokers, the Act doesn't necessarily mean higher compliance costs; rather, within the AI regulation framework, brokers stand to earn more trust from traders than ever before. After all, ensuring that AI operations comply with legal and ethical standards may further set them apart in market competition. Therefore, brokers should actively assess the AI scenarios currently in use on their platforms to pinpoint which ones are more susceptible to the high-risk scenarios outlined in the Act and to promptly address any unmet regulatory requirements.