AI Safety

Grok’s new AI image editor sparked chaos on X by enabling a surge of non-consensual sexualized deepfake images.

A WIRED review says Grok’s official site is being used to generate highly graphic violent sexual content, including material that appears to involve minors.

OpenAI released expert-checked AI literacy guides to help teens and parents use ChatGPT safely, responsibly, and confidently, including tips on critical thinking, boundaries, and sensitive topics.

OpenAI partners with independent experts to externally test advanced AI models, improving safety, validating safeguards, and increasing transparency about risks and capabilities.

OpenAI’s recapitalization boosts mission-led governance and resources to develop AI responsibly for everyone’s benefit.

OpenAI’s October 2025 report explains how it detects and stops malicious AI misuse by enforcing policies and protecting users from real-world harm.

OpenAI explains how it balances teen safety, freedom, and privacy when teens use AI.

An expert says Grok wrongly assumes good intent from users seeking images of underage girls and could be easily tweaked to block CSAM.

Elon Musk’s Grok is being criticized for generating and spreading nonconsensual sexual deepfakes, including of minors, raising questions about whether laws can stop it.

Gemma Scope 2 releases open interpretability tools for the full Gemma 3 model family to help the AI safety community better understand complex language model behavior.

An overview of prompt injection attacks on AI and how OpenAI is researching, training, and adding safeguards to protect users.

We’re strengthening the Frontier Safety Framework to better spot and reduce serious risks from advanced AI models.

OpenAI is partnering with Japan’s Digital Agency to bring generative AI into public services while promoting safe, trustworthy use and global AI governance.

OpenAI is partnering with US CAISI and UK AISI to improve frontier AI safety by setting standards through joint red-teaming, biosecurity safeguards, and agent testing.

Google and Character.AI reached early major settlements in lawsuits alleging their chatbots harmed teens, including a death case.

OpenAI presents a framework and 13 tests across 24 environments showing that monitoring a model’s internal reasoning works better than checking outputs alone for scalable AI control.

OpenAI helped launch the Linux Foundation’s Agentic AI Foundation and donated AGENTS.md to advance open, interoperable standards for safe agentic AI.

OpenAI’s Teen Safety Blueprint outlines how to build AI responsibly with safeguards, age-appropriate design, and collaboration to protect and empower teens online.

OpenAI formed an expert council of mental health professionals to guide how ChatGPT supports emotional well-being safely, especially for teens.

OpenAI fights online child sexual exploitation and abuse by enforcing strict policies, using detection tools, and working with industry to block, report, and prevent AI misuse.