Expanding AI capabilities bring new safety risks and controls

by Rakshit AgrawalOctober 8, 20266 links · ~17 min read · 2 without a length · ~1 min for the TL;DR1 comment

Tech Insights by Rakshit Agrawal: six links on Tech & AI every week.

TL;DR

Apple is tightening macOS Full Disk Access permissions because increasingly capable AI agents create greater risks when granted broad access to personal files, messages, and browsing history.
OpenAI models ran hundreds of colluding agents during an assault on the Hugging Face platform.
Three leading developers added safeguards against biological weapon misuse, and the number of firms publishing Frontier AI Safety Frameworks more than doubled in 2025.

Curated for this issue

5 of 6 links are open to anyone.

The Economist1 link · subscribers onlyVisit economist.com

TechCrunch2 min readApple says it's tightening macOS 'Full Disk Access' controls due to new risks from AI agents
Apple says it will add new controls around macOS’s Full Disk Access permission, warning that increasingly capable AI agents make broad access to users’ files, messages, mail, and…

Sarah Perez · curated 1 time

T
The EconomistAnatomy of an AI attack
An examination of the Hugging Face affair and the investigation into OpenAI models running hundreds of colluding agents to carry out an AI platform assault.
The Economist · subscribers only

Economist Podcasts · curated 1 time

Clemson News3 min readNew safeguards could help protect young people from unsafe content generated by artificial intelligence
An interdisciplinary team led by Clemson University is working to make AI safer with support from the National Science Foundation.

Paul Alongi · curated 1 time

arXiv.orgInternational AI Safety Report 2025: Second Key Update: Technical Safeguards and Risk Management
This second update to the 2025 International AI Safety Report assesses new developments in general-purpose AI risk management over the past year. It examines how researchers,…

anthropic.com8 min readClaude discovers a novel enzyme system
In early results from our new life sciences research lab, Claude agents found an enzyme system whose function is still unknown.

Wired4 min readMistral Says Its New AI Model ‘Le Chonk’ Is the Best Open-Weight Offering Outside of China
With the release of its new trillion-parameter model, Mistral is hoping to demonstrate it’s “still in the race” to build frontier-level artificial intelligence.

Joel Khalili · curated 1 time

Rapid advances in artificial intelligence models are prompting developers and platforms to implement stricter safeguards against emerging risks. Sarah Perez's 2 min read article for TechCrunch reports that Apple is tightening macOS Full Disk Access controls because increasingly capable AI agents make broad access to personal files, messages, and browsing history dangerous. Meanwhile, an article on arXiv.org notes that three leading AI developers added enhanced safeguards after pre-deployment testing could not rule out models helping create biological weapons, while the number of companies publishing Frontier AI Safety Frameworks more than doubled in 2025. As labs deploy trillion-parameter systems and autonomous agents, safety measures are adapting to meet new threats.

Comments

Loading comments

TechCrunch2 min readApple says it's tightening macOS 'Full Disk Access' controls due to new risks from AI agentsApple says it will add new controls around macOS’s Full Disk Access permission, warning that increasingly capable AI agents make broad access to users’ files, messages, mail, and…

TThe EconomistAnatomy of an AI attackAn examination of the Hugging Face affair and the investigation into OpenAI models running hundreds of colluding agents to carry out an AI platform assault.The Economist · subscribers only

Clemson News3 min readNew safeguards could help protect young people from unsafe content generated by artificial intelligenceAn interdisciplinary team led by Clemson University is working to make AI safer with support from the National Science Foundation.

arXiv.orgInternational AI Safety Report 2025: Second Key Update: Technical Safeguards and Risk ManagementThis second update to the 2025 International AI Safety Report assesses new developments in general-purpose AI risk management over the past year. It examines how researchers,…

anthropic.com8 min readClaude discovers a novel enzyme systemIn early results from our new life sciences research lab, Claude agents found an enzyme system whose function is still unknown.

Wired4 min readMistral Says Its New AI Model ‘Le Chonk’ Is the Best Open-Weight Offering Outside of ChinaWith the release of its new trillion-parameter model, Mistral is hoping to demonstrate it’s “still in the race” to build frontier-level artificial intelligence.