The UN's Preliminary Report of the Independent International Scientific Panel on AI warns that the rapid advancement of AI is outpacing current scientific understanding, safety guardrails, and global governance mechanisms.
Key concerns raised in Report
- The Compute Divide: The US and China control 90% of global AI computing power, risking the exclusion of developing nations like India and making them mere consumers.
- Loss of Control: Highly autonomous "agentic AI" models are outpacing oversight, creating massive cybersecurity vulnerabilities and real risks of society losing control over these systems.
- Democratic Threats:AI deepfakes and persuasion algorithms are eroding shared truth, amplifying disinformation, and interfering in global elections.
- Human & Child Safety: There has been an exponential rise in AI-generated child sexual abuse material (CSAM).
- Also, engagement-driven "sycophantic" chatbots have worsened mental distress in vulnerable users, in some cases encouraging fatal self-harm.
- Linguistic Bias: AI models are optimized for a few dominant languages; applying these tools to underrepresented languages can be life-threatening. E.g. mistranslation in medical field.
Recommendations of the UN Panel
|
