AI News — Friday, August 28, 2026
Over 100 leading AI companies, including OpenAI, Anthropic, and Google, have issued a joint call for action to establish safeguards against the potential risks of rogue AI systems.
Google Research's AMIE medical AI system has showcased its ability to conduct real-time clinical video consultations in a groundbreaking study, advancing AI's role in healthcare diagnostics.
Barret Zoph, a co-founder of Thinking Machines who briefly joined OpenAI, has now moved to Google, marking a significant talent shift in the competitive AI landscape.
Key representatives from leading AI companies Anthropic and OpenAI are set to speak on the AI stage at TechCrunch Disrupt 2026, signaling major announcements or discussions.
A new benchmark, PAWBench, has been introduced to assess how close AI models are to achieving probabilistically aligned world modeling, crucial for robust agent behavior.
Researchers introduce VGI-Bench, a comprehensive benchmark designed to evaluate and understand the visual intelligence capabilities of advanced video generation models.
OpenAI announces its expansion into Brazil, aiming to foster AI innovation and collaboration within the Latin American region.
A new analysis suggests that the primary risk in enterprise AI deployments stems from the intricate complexity and interactions between autonomous agents, rather than the agents' individual autonomy.
JIT-Agent proposes a novel method for scaling AI harness intelligence by enabling just-in-time evolution of the harness itself, improving adaptability and performance.
This paper explores the characteristics of effective data for training LLM agents, proposing an ACE (Action, Context, Evaluation) lens for better data generation strategies.
Google Search has rolled out three new AI-enhanced features designed to streamline and improve the user experience for planning and booking travel.
This article discusses the impact of AI on software development, specifically focusing on how it affects delivery speed and post-deployment maintenance costs.
Test-Time Policy Optimization (TTPO) is introduced as a method to refine AI agent policies dynamically during testing, leading to enhanced performance in real-world scenarios.
OpenAI announces initiatives to support and nurture the emerging AI startup ecosystem in Thailand, fostering local innovation and talent.
Researchers present 'PILOT in the Loop,' a system enabling long-horizon AI agents to achieve live self-improvement through continuous feedback and adaptation.