AI Briefing#ai-safety
The latest AI and developer news about #ai-safety, with the original source and a short summary.
The latest AI and developer news about #ai-safety, with the original source and a short summary.
OpenAI has released a prompt-based safety policy package to help developers build safe AI services for teens.
OpenAI announced multi-faceted safeguards for safe video generation on Sora 2 and the Sora app.
OpenAI has built a monitoring system based on GPT-5.4 Thinking to detect misalignment risks in its internal coding agent.
OpenAI Japan has announced the 'Japan Teen Safety Blueprint' to support teenagers' safe use of generative AI.
Anthropic has launched The Anthropic Institute to address the social and economic risks of frontier AI.
OpenAI released IH-Challenge, a reinforcement learning dataset that helps models correctly judge priority among conflicting instructions.
Current reasoning models have difficulty intentionally manipulating their chain-of-thought (CoT), which is a positive signal for maintaining AI safety monitoring.
OpenAI has released the GPT-5.4 Thinking system card, which applies cybersecurity risk mitigation measures for the first time.
OpenAI signed an AI system deployment agreement with the U.S. Department of War, specifying strong safety guardrails such as a ban on domestic surveillance and autonomous weapons use.
OpenAI is strengthening ChatGPT's mental health-related safety features and has stated its position on related lawsuits.
Anthropic has released the third version of the 'Responsible Scaling Policy (RSP),' a voluntary framework for mitigating the catastrophic risks of AI.
OpenAI is funding UK AISI's independent research fund 'The Alignment Project' with $7.5 million to help secure the safety of AI alignment.
OpenAI is introducing Lockdown Mode and an 'Elevated Risk' label to ChatGPT to prevent prompt injection attacks.
Anthropic is donating $20 million to Public First Action to strengthen AI governance and secure US AI leadership.
Stability AI has officially joined the Tech Coalition, a global alliance dedicated to preventing online child sexual exploitation and abuse.
OpenAI is pushing to build localized AI systems that reflect each country's language, laws, and culture so that everyone can benefit.
OpenAI has released GPT-5.3-Codex, an agentic coding model that combines reasoning ability with coding performance.
Anthropic has strengthened model training and product-level safeguards so that Claude responds appropriately to sensitive conversations such as suicide and self-harm.
OpenAI has revealed the operating principles and recommendation algorithm behind the Sora feed, aimed at fostering creativity and building a safe environment.
Hallucination, bias, jailbreaking, and sycophancy in generative AI are threatening safety and trust.
A support program worth €500,000 is being run to support AI safety and wellbeing for youth in the EMEA region.
Anthropic strengthened Claude's safety and helpfulness through Constitutional AI, which is based on explicit principles.
Mariano-Florentino Cuéllar has been appointed as a new member of Anthropic's Long-Term Benefit Trust, which supports the company's public benefit mission.
OpenAI is introducing a feature that predicts ChatGPT users' age to strengthen protections for teens.