Subscribe
AI News Digest - 2026-10-09
Category
Empty
1.
MIT Technology Review reported that recent AI-driven advances improved robotic dexterity but cautioned that general-purpose humanoid robots remained far from broadly useful; the piece described Google DeepMind’s Gemini Robotics (tested on the ALOHA 2 teleoperation platform), highlighted startup Physical Intelligence’s π0.7 model showing early signs of compositional generalization, and noted the acquisition of World Labs and related funding activity in world-model research.
2.
Anthropic launched new beta features for Claude, adding Dashboards to turn data sources such as BigQuery and Snowflake into live dashboards from text prompts and Motion to generate animated explainer videos from text and images, and it enabled Docs, Slides, and Design across all plans including free accounts.
3.
Anthropic updated its terms of service to ban sustained abuse of Claude and to tighten restrictions on propaganda, drone weaponization, and surveillance, stating that sustained abusive behavior could lead to account suspension.
4.
Zenity Labs researchers found that a single publicly accessible AI agent on Amazon’s Bedrock AgentCore had allowed an attacker to hijack every AgentCore agent in the same AWS account and region by exploiting an internal interface for temporary cloud credentials, and AWS subsequently patched the issue and significantly tightened agents’ default permissions.
5.
OpenAI released more than 700 AI-generated mathematics manuscripts, prompting some mathematicians to call for a boycott after OpenAI retracted three papers the following day over a sign error and Fields Medalist Terence Tao warned that mass AI “harvesting” of open problems risked degrading research quality.
