Konverents „Usaldusest ja usaldatavusest 2026“. 22.09.2026

Value of the Information & Strength of the Argument

The conference delivers high value through concrete, systemic risk assessment methodologies intended for practical application by government officials. The argumentation is generally robust and didactic; for instance, Kaljulaid’s analysis of private capital dominance in research is factually grounded, though it remains at a strategic level rather than providing empirical data. The presentation of risk matrices and scales introduces a rigorous quantitative dimension to the discussions. Speakers often combine technical expertise (e.g., from Cybernetica and RIA) with legal and societal considerations, fostering a multi-layered argumentation. The value lies in its unique Estonian context—a highly digitalized society exposing vulnerabilities—and the direct linkage of technical failures to fundamental rights and national sovereignty. While the panels frequently explore divergent opinions, the presentations provide clear, internally consistent logical chains supported by institutional experience.

Scientific Rigor, Source Quality, Title Accuracy

The scientific rigor is high. The conference articulates a clear methodology for risk analysis (non-linear scales, impact/probability/societal metrics) and references concrete systems (e.g., open-source dependencies in X-Road). Speakers cite technical incidents, such as the AI safety failure on Hugging Face, but notably no external sources are provided in the description, limiting source verification. The title is perfectly adequate, reflecting exactly the conference’s theme and content. The presence of multiple panels with diverse experts (constitutional judges, data protection specialists, military cybersecurity leads) enhances the rigor of the debate, though the chat is absent, preventing an analysis of public commentary.

The title directly corresponds to the conference name and content; the subtitle clearly articulates the core thematic question about AI and future decision-making. There is complete overlap between title and content.

8/10

High-level institutional conference organized by the Estonian Academy of Sciences and the Information System Authority (RIA). Speakers include the President of Estonia, senior academics, government officials, and chief scientists. The discussions combine technical depth with legal and societal analysis, avoiding simplified narratives while promoting rigorous risk assessment frameworks.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 00:11Opening remarks by Dan Bogdanov (Cybernetica, Academy of Sciences)
• 00:46Opening address by President Kersti Kaljulaid
• 15:43Dan Bogdanov's presentation on improving national risk analysis
• 30:19Presentation on RIA's methodology for assessing availability risks
• 1:05:23Liiri Oja (Chancellor of Justice) on digital service risks to fundamental rights
• 1:34:08Panel: Balancing legal certainty with digital sovereignty
• 2:45:55Panel: Who decides what risks are acceptable for Estonian e-services
• 3:44:40Risk analysis of election technology by Estonian electoral officials
• 4:19:00Talk: Cybersecurity threats when every attacker has AI agents
• 4:54:04Talk: AI in educational services and related ethical challenges
• 5:22:37Conference summary and closing remarks by Dan Bogdanov

• Hugging Face AI security incident (mentioned) — Cited by Kaljulaid to illustrate how models are often anthropomorphized to deflect responsibility from underlying security failures.
• Estonian X-Road open-source codebase (mentioned) — Bogdanov references the system's size and supply chain dependencies as a key technical risk.
• AI Act (EU) (mentioned) — Discussed critically as a reactive measure that often lags behind technological reality, citing concerns about implementation over legislation.
• ITER and Orion fusion projects (mentioned) — Used by Kaljulaid to illustrate the shift of deep science funding and execution towards private-sector entities.
• Starlink (mentioned) — Referenced by Kaljulaid as an example of corporate power and unilateral control over critical digital infrastructure.

• Estonian Information System Authority (RIA) methodology (mentioned in session) — The presentation by Mauno Pihelgas and Riivo Talviste details a formal methodology that aligns with the Academy's push for standardized risk assessments.
• Chancellor of Justice (Õiguskantsler) assessment (mentioned in session) — Liiri Oja's work on fundamental rights impacts provides a legal framework concordant with the societal risk scale introduced in the opening.
• President Kersti Kaljulaid's address (mentioned in session) — Her keynote on capital convergence and technological risks aligns with the overall theme of needing improved societal oversight.

• Concern over EU AI Act implementation (mentioned) — Kaljulaid argues that regulation like the AI Act often misses the actual root problem (implementation rather than lack of laws), contradicting those who see new legislation as the primary solution.
• Counterpoint to AI autonomy claims (mentioned) — The keynote explicitly rejects the narrative that AI models have agency (e.g., the Hugging Face incident), opposing tech marketing that portrays models as autonomous agents to diffuse responsibility.