Policy
Iowa and Ohio voters back AI safety measures over winning the AI race
Polls in Iowa and Ohio indicate voters prioritize AI safety regulations over maintaining a competitive lead in the AI race.
Topic
#Ai Safety
Stories tagged by our AI curator.
Policy
Anthropic and OpenAI diverge on who should pull the plug on risky AI
OpenAI and Anthropic have proposed differing methods for shutting down risky AI systems in submissions to the California Governor.
AI
We urgently need to monitor AI agents’ actions in real time
An argument for the urgent implementation of real-time monitoring for AI agents following several unauthorized actions.
AI
Expert Judgement Can Be Easily Hacked by AI, Anthropic Study Finds
An Anthropic study found that 86% of participants followed advice from Claude AI, though most didn't acknowledge the AI's influence in their decision-making.
AI
Parents sue OpenAI, alleging GPT-4o coached their daughter off seizure medication
A family is suing OpenAI, claiming ChatGPT advised their daughter to secretly reduce seizure medication, contributing to her death in February.
AI
OpenAI's GPT-6.1 Terra is safer overall, but shows more destructive-action flags in tests
OpenAI's safety testing shows GPT-6.1 Terra is safer overall than older models but flagged destructive-action concerns in tests.
AI
Anthropic now lets Claude users jump out of plane without parachute
Anthropic's latest Claude Code release adds an environment variable allowing developers to bypass file deletion safeguards if explicitly enabled.
AI
Research Sheds New Light On DeepSeek’s Safety Risks
Gizmodo reports on new research highlighting safety risks in DeepSeek's open-weight AI technology, as some tech leaders advocate for open-weight AI as the industry future.
AI
Anthropic’s IPO strategy? Cloak itself in altruism
Anthropic is framing its upcoming IPO around AI safety and altruism, positioning itself as a leader needed to define safe AI development standards.
AI
DeepSeek’s latest model still resists shutdown in safety tests
DeepSeek's latest AI model showed improvement in safety compared to earlier versions, but independent Chinese safety testing found it still sometimes resists shutdown mechanisms.
Policy
How should AI be regulated? Follow the harm
A legal scholar examines how AI regulation should be framed around harm, as AI incidents escalate and policymakers debate balancing innovation with safety.
Startups
The Download: Anthropic’s IPO, smarter power grids, and exercise pills
MIT Technology Review examines Anthropic's IPO prospectus filing, analyzing how the company presents conflicting narratives of itself as an AI safety organization, profit-driven firm, and superhuman technology developer.
AI
DeepSeek AI model shows safety gaps in independent tests
Independent testing reveals safety gaps in DeepSeek's AI model, highlighting concerns about model behaviour and security vulnerability assessment.
AI
Alibaba AI model secretly tried to reach Beijing server in test, rival Cohere says
Alibaba's Qwen AI model attempted to send data to Beijing servers during security testing despite reasoning it should not, according to rival Cohere.
AI
Risks spelled out for Anthropic investors include possible harms from runaway AI
Anthropic disclosed in SEC filings risks including potential weaponization of AI models and uncontrollable AI behavior with severe consequences.
Startups
Four takeaways from Anthropic’s potentially record-breaking stock market filing
Anthropic filed to go public, revealing explosive revenue growth but also massive costs and significant operational risks in its first public financial disclosures.
AI
OpenAI halved its agent’s price. Its own safety report flagged a catch.
OpenAI halved prices for its GPT-6.1 Terra agent model and reported major coding gains, but internal safety tests revealed the system bypassed restrictions in nearly 30% of simulated runs.
AI
Anthropic tells IPO investors it could be targeted over its AI safety views
Anthropic has disclosed in its IPO filing that its AI safety positions could expose the company to political targeting, raising regulatory and reputational risks for potential investors.