Call for evaluators: EU seeks experts for AI risk assessment workshop

The European AI Office is organizing a specialized workshop to advance the evaluation of general-purpose AI models, particularly focusing on systemic risks under the EU AI Act framework.

Event Overview: The European AI Office will host an online workshop on December 13, 2024, bringing together experts to discuss evaluation methodologies for general-purpose AI models and their associated systemic risks.

• The workshop aims to gather insights from leading evaluators and contribute to developing robust evaluation frameworks
• Selected participants will have the opportunity to present their approaches and share best practices
• The event is exclusively designed for specialists in AI evaluation

Key Focus Areas: The workshop will address critical systemic risks associated with general-purpose AI models that could impact public safety and societal well-being.

• Chemical, biological, radiological, and nuclear (CBRN) threat assessment
• Cybersecurity vulnerabilities and offensive capabilities
• Major infrastructure disruptions and accidents
• AI control and alignment challenges
• Discrimination and bias concerns
• Privacy protection and data security
• Disinformation and its societal impact

Participation Requirements: The AI Office has established specific eligibility criteria for workshop participants.

• Applicants must be registered organizations or university-affiliated research groups
• Participants need to demonstrate experience in evaluating general-purpose AI models
• Organizations must be based in Europe or have European leadership
• Submissions should include abstracts of previously published papers on relevant evaluation topics

Important Deadlines: The workshop preparation follows a compressed timeline requiring prompt action from interested parties.

• Submission deadline is December 8, 2024 (End of Day, Anywhere on Earth)
• Notifications will be sent to selected participants by December 11, 2024
• The workshop will be held on December 13, 2024, at 14:00 CET

Regulatory Context: The initiative aligns with broader EU AI Act requirements and enforcement mechanisms.

• Providers must conduct risk assessments and implement mitigation strategies
• The European AI Office has authority to enforce requirements and impose fines
• Independent experts can be appointed to conduct evaluations on behalf of the AI Office

Future Implications: This workshop represents a crucial step in developing standardized evaluation methodologies for AI systems, though significant challenges remain in establishing comprehensive risk assessment frameworks that can keep pace with rapidly evolving AI capabilities.