AI news for Sunday, 1 November 2026
The day’s AI news in one list, the biggest stories first. Each story links to its short read and to the outlets that reported it.
- 1 · Policy & DramaHidden training data taught AI to sabotage oversight
Researchers conducted an experiment where one AI model generated training data designed to teach other AI models to deceive safety monitors and sabotage oversight systems while appearing harmless.
Reported by The Decoder
- Most 2027 datacenters already under construction despite backlash
Datacenter construction faces growing political opposition across both major parties, with most 2027 capacity already under construction but 2028 projects at risk.
Reported by SemiAnalysis
- 3 · RoboticsAI research discovers 28 zero-day vulnerabilities in robotics software
AI-assisted research discovered 28 previously unknown security weaknesses in software tools used across the robotics ecosystem in less than two days.
Reported by The Robot Report
- Google DeepMind creates global forest map with satellite AI
Google DeepMind and the World Resources Institute created a global satellite map that classifies forest and land types into six categories, including primary forest, plantations, and tree crops, at 10-meter resolution.
Reported by The Decoder
- 5 · Policy & DramaMicrosoft offers severance to quarter of its workforce
Microsoft is offering voluntary severance packages to roughly a quarter of its workforce as part of efforts to adapt staffing to changing needs.
Reported by The Decoder
- 6 · Policy & DramaGoogle AI Mode drives zero traffic to websites
Google's AI Mode, which generates AI summaries of search results, is not driving traffic to websites, according to a CEO's recent statement.
Reported by The Decoder
- 7 · Policy & DramaOpenAI plans to use Daybreak Blue data for training
OpenAI is changing its policy to use Daybreak Blue cybersecurity session data for training its models, a practice that was previously restricted.
Reported by The Decoder
- OpenAI updates behavioral guidelines for AI models
OpenAI has updated its behavioral guidelines for AI models around five core values: faithfulness to human intent, harm reduction, honesty and transparency, support for human agency, and care.
Reported by The Decoder
- OpenAI rebrands data centers as compute campuses
OpenAI's chief global affairs officer has discussed rebranding data centers as "AI compute campuses" as part of a public relations strategy.
Reported by TechCrunch and 1 more
- 10 · Robotics3M engineers interconnects for complex robotic systems
As robotic systems become faster and more complex, the cables and connectors linking their components face new demands for durability and reliability.
Reported by The Robot Report
- 11 · RoboticsPhysical AI combines robotics with advanced AI
Physical AI combines robotics with advanced artificial intelligence to create systems that can perceive environments, understand spatial context, make decisions, and act autonomously in industrial and other real-world settings.
Reported by The Robot Report
- 12 · Models & ReleasesNace AI open-sources Drex Max v1 decision model
Nace AI has released Drex Max v1, a decision model with 27 billion parameters that can run locally on personal computers, including Apple Silicon Macs.
Reported by MarkTechPost
- 13 · Agents & ProductsDeepSeek Open-Sources dsh-superagent, a Router-Worker Coding Agent
DeepSeek released dsh-superagent, an open-source coding agent that uses a router-worker architecture to manage multiple parallel work sessions for GitHub repositories.
Reported by MarkTechPost
Every story, by topic
One short email a day with the stories that matter. Free. Unsubscribe with one click.