Big news at the UN

Substack · Watch/Read · (gathered 2026-09-23)

• Gary Marcus, speaking at the UNGA, advocated for a sober, practical approach to AI regulation, rejecting both “zero regulation” and “doomer” extremes, and focusing on immediate threats like deepfaked disinformation and large-scale cyberattacks rather than superintelligence or extinction.
• He highlighted the need for increased reliability, robust cybersecurity, and holding AI companies accountable for harms, citing the WannaCry virus as a cautionary tale and criticizing OpenAI for releasing models that are harder to monitor.
• Marcus proposed three immediate steps: establishing international advisory commissions for pre-release evaluation and ongoing post-release auditing (with authority to demand change or recall), and an international agreement to prohibit the deployment of demonstrably harmful AI architectures.