Google Research
·
1 month ago
Researchers introduced Regularized f-Divergence Kernel Tests, a new auditing framework presented at AISTATS 2026 for verifying machine unlearning in AI models. The framework uses multiple divergence measures to detect whether unlearned models successfully removed specific training data, with the hockey-stick divergence test detecting privacy violations using only thousands of samples compared to millions required by previous methods. This enables auditors to mathematically prove privacy compliance with minimal manual tuning and fewer data samples, addressing regulatory requirements like GDPR's Right to be Forgotten.
Google DeepMind
·
1 month ago
Google DeepMind, Schmidt Sciences, the Cooperative AI Foundation, and ARIA announced $10 million in research funding to study safety risks that emerge when multiple AI agents built by different organizations interact with each other. The application deadline is August 8, 2026, with results announced in Autumn 2026. The funding aims to develop frameworks for understanding and controlling unpredictable collective behaviors that arise in large-scale multi-agent systems, which existing safety evaluations conducted on individual models cannot address.
Platformer
·
1 month ago
Molly Kinder, who has spent three years researching how generative AI affects work, argues that job losses will concentrate in high-paid white-collar knowledge jobs during a "messy middle" period lasting potentially decades. Unlike blue-collar or service-sector jobs, computer-based work in law, finance, consulting, and accounting face the greatest near-term exposure to language models like ChatGPT. Rather than adopting universal basic income, Kinder proposes targeted interventions including workforce reinvestment funds, wage insurance, and public job creation to manage the transition without destroying labor market incentives.
