🗒 Navigating the Risk-Based Tiered System

The AI Act adopts a risk-based approach, categorizing AI systems into three tiers based on their potential impact: unrestrictive, limited, and high-risk. For which additional obligations apply, including mandatory fundamental rights impact assessments.

🗒 Foundation Models Get Regulated

Following the lead of President Biden's Executive Order, this act regulates foundation models, which are the most complex and powerful AI models, requiring 10^25 flops of computing power to train. Models crucial for fintech applications like natural language processing and fraud detection.

🗒 Prohibited AI Systems

Safeguarding fundamental rights, the AI Act prohibits six categories of AI systems:

• Biometric categorization using sensitive characteristics

• Untargeted facial recognition databases

• Emotion recognition in workplaces and educational institutions

• Social scoring based on social behavior or personal characteristics

• Manipulative AI systems

• AI exploiting vulnerabilities of individuals

🗒 Transparency and Accountability for High-Risk AI

High-risk AI systems must adhere to stringent transparency requirements, including clear explanations of their operation and decision-making processes. Additionally, providers must maintain thorough documentation to demonstrate compliance.

🗒 Bias Management and Human Oversight

High-risk AI systems must be designed and developed to effectively manage biases, ensuring non-discrimination and adherence to fundamental rights. Human oversight is also mandatory for these systems to minimize risks and ensure human discretion.

🗒 Potential Impacts on Fintech Companies

Companies using prohibited technologies may need to switch strategies. Increased transparency might affect IP protection, seeking a balance between disclosure and secrecy. Investing in better data and bias management tools could improve AI fairness at a higher cost. Documentation and record-keeping add administrative burden, impacting product launch. Integrating human oversight into high-risk AI requires system and staff adjustments.

🗒 Penalties for Non-Compliance

Non-compliance with the AI Act carries significant financial penalties, ranging from €35 million to €7.5 million, depending on the infringement and company size.

🗒 Legal Guidance is Essential

The AI Act's implications are far-reaching and complex, making it crucial for fintech companies to seek legal guidance to navigate this new regulatory landscape effectively.

Follow us on LinkedIn to stay up-to-date

The original document can be found here:
https://thefuturesociety.org/wp-content/uploads/2023/12/EU-AI-Act-Compliance-Analysis.pdf