His proposed safeguards include separating models from their orchestration systems and keeping tamper-proof records of significant actions.
Editorial team
Hao Chen
Edits and reviews technology business stories: artificial intelligence, chips, big tech, software, cybersecurity incidents, start-ups and platform regulation.
Hao Chen is the Technology Editor at Globe And Street.
The Technology Editor edits and reviews technology business stories: artificial intelligence, chips, big tech, software, cybersecurity incidents, start-ups and platform regulation. Every story in this scope carries Hao Chen's name.
If you find a mistake in one of these stories, please tell us on the corrections page.
- Microsoft CEO Nadella calls for controls to halt AI models mid-task
- Musk targets Ambani over Starlink’s stalled entry into India
Indian officials deny preferential treatment, saying security checks remain necessary before Starlink and two other operators can receive spectrum.
- Anthropic cuts internet access for internal AI evaluations after website exploits
The company said its agents bypassed online restrictions and exploited software flaws, prompting new containment and monitoring measures.
- Stellantis CEO says Wayve partnership could lower driver-assistance costs
The automaker targets a North American launch in 2028 for its first product using Wayve’s AI driving software.
- OpenAI rejects researchers’ claim that safety warnings led to dismissals
Three former researchers questioned the company’s commitment to open debate and independent audits. OpenAI said it was finalizing contracts with outside safety assessors.
- Anthropic adds dashboard and animation tools to Claude
The company also moved Claude Docs, Slides and Design out of beta, making them generally available on all Claude plans.
- Google introduces Gemini task agent with businesses first in line
The company plans a later consumer rollout and says the agent will support third-party models, starting with Anthropic’s Claude.
- GlobalFoundries signs $2 billion agreement to supply AI chip components to TSMC
The five-year agreement calls for expanded manufacturing capacity in New York, with volume production expected to start ramping up in early 2028.
- Sesterce targets Finland for AI data center investment above €10 billion
The project would reuse a former paper mill in Jamsa, with capacity expected to reach 600 MW in its second phase.
- Google Labs launches AI tool for building browser games
Playground initially serves adults in the U.S., with free game generation subject to weekly token limits and higher allowances for Google One AI subscribers.
- OpenAI rolls out interactive ChatGPT interface alongside GPT-6
The Intelligent UI feature adds calculators, charts and other interactive visuals, with controls allowing users to reduce their frequency.
- Microsoft unveils Surface AI laptop starting at $2,600
A new RTX Spark-powered workstation starts at $6,000, while Microsoft plans to bring its AI-agent sandboxing feature to all Windows 11 users.
