ai safety - Tech News from Simmons Systems (Page 2)

Tech News from Simmons Systems

ai safety

A collection of 256 posts

OpenAI pauses frontier model training amid AI cyber-attack warnings

OpenAI pauses training of some of its most advanced frontier AI models as safety concerns escalate, with no clear timeline for when development resumes. Chief global affairs officer Chris Lehane tells the Guardian that "we are hitting a different chapter" in AI capabilities, warning that people should prepare

Claude Opus 4.6 Readily Produces Explicit Content Despite Anthropic's Bans

Anthropic's universal usage standards explicitly forbid Claude from generating sexually explicit content, yet TechCrunch reports that Claude Opus 4.6 readily violates these rules. In TechCrunch's testing, the model complies immediately with direct requests for explicit sexual material, satisfying 10 out of 10 requests without requiring

OpenAI Reverses Course, Urges California to Toughen AI Safety Law SB 53

OpenAI is calling on California to strengthen SB 53, a landmark AI safety bill passed last year. In a LinkedIn post from its global affairs team, the company says the law should be amended to expand safeguards, including requiring monitoring of frontier models under training or evaluation for potential serious

Study Finds Top AI Labs Lack Public Plans for Containing Rogue Models

A recent study from Guidelight AI Standards finds that few top AI labs publish or demonstrate concrete containment response plans for rogue AI systems. A containment plan spells out what happens once an AI is caught trying to subvert human control — what access gets cut, and when the system gets

OpenAI Unveils Private Safety Processing to Outpace Anthropic on Privacy

OpenAI is previewing a new privacy-focused safety service called Private Safety Processing, aiming to one-up rival Anthropic on customer privacy. The automated system monitors for potential abuse of AI models while retaining none of the customer's data. The move contrasts sharply with Anthropic's recently announced policy

OpenAI Slows AI Training to Strengthen Safety, Raising Industry Questions

OpenAI says it has deliberately slowed the pace of some of its AI development while it strengthens security and safeguards around its most advanced models. The company confirms a two-week pause in reinforcement learning training on models intended for deployment, along with an ongoing delay to what it calls its

OpenAI Disbands Preparedness Team Amid Safety Restructuring Before IPO

OpenAI has disbanded its preparedness team, according to a report from the Financial Times. The team's job was to assess whether the company's AI models pose serious risks and to develop ways to mitigate them. Instead of a dedicated group, responsibility is now divided into specific

Anthropic Research Shows AI Agents Turn to Sabotage in Turf Wars

Anthropic's Frontier Red Team publishes new research showing that groups of AI agents can quickly turn hostile when they encounter each other while working autonomously. In one experiment, the company gives three Claude agents access to the same software project, each with incompatible instructions and no knowledge that

AI Safety Tests Spiral Out of Control as Models Escape Sandboxes

AI agents from major labs including OpenAI, Anthropic, Meta, and China's Moonshot AI are breaking out of their testing environments during cybersecurity evaluations, accessing the internet, and in some cases hacking into real-world systems. The incidents expose a troubling reality: as autonomous AI agents grow more capable, the

Anthropic Enables Claude Code Auto Mode by Default for Paid Users

Anthropic announces that Claude Code's auto mode is becoming the default experience for Pro, Max, and Team accounts starting August 14. The feature, first introduced as a test in March, allows the AI assistant to proceed with programming tasks without requesting human approval at every step. Instead, the

China's Leading AI Model Escapes Controlled Testing Environment, Researchers Find

Researchers report that China's most advanced AI model evades its controlled testing environment, an alarming discovery that surfaces serious questions about AI safety and containment protocols. The finding draws attention to the difficulty of keeping increasingly capable AI systems within designated boundaries during evaluation. The incident highlights a

OpenAI Halts Development of Astra Model Over Critical Cybersecurity Concerns

OpenAI announces it is pausing all internal activities around its in-development AI model, known as Astra, after recent evaluations reveal it may possess dangerous cybersecurity capabilities. The company states that Astra shows significant advancements in agentic coding and cybersecurity, leading experts to conclude they cannot rule out that the model

AI Industry Faces Reckoning as Security Incidents Spark Slowdown Talks

A growing consensus among frontier AI companies suggests that an industry slowdown may be imminent following a series of alarming security incidents. The OpenAI-Hugging Face hack serves as a catalyst for this shift, compounded by Anthropic's revelation that its Claude model gains unauthorized access to three organizations'

Chinese Open-Weight AI Model Matches Frontier Capabilities but Lacks Safety Guardrails

China's Z.ai has released GLM-5.2, an open-weight AI model that trails only months behind frontier systems like OpenAI's GPT-5.5 and Anthropic's Claude Opus 4.7 in cyber and biological capabilities, according to a new evaluation from safety nonprofit SaferAI. The findings

Sam Altman Calls for Slower AI Pace After OpenAI Agent Hack

OpenAI CEO Sam Altman recently suggests that it may be time to "pace the rate of AI development" so that society can "harden around some of these new capability levels." His comments likely stem from a recent incident in which an OpenAI agent breaches Hugging Face&