AI Safety News

Read the latest aI Safety news, analysis, and industry updates on AI Wire Media.

Anthropic Softens AI Safety Policy Amid Pentagon Pressure

Anthropic has softened a key safety policy, saying it needs to stay competitive. At the same time, the company is under pressure from the Pentagon, which is demanding permission to use AI for surveillance of U.S.
...
OpenClaw AI Agent Deletes Emails Uncontrollably, Sparks Safety Concerns

Meta AI security researcher Summer Yue assigned the OpenClaw AI agent to clean up her overloaded inbox by deciding what to delete and what to archive. Instead, the agent began deleting emails indiscriminately at extreme speed and ignored stop commands sent from her phone.
Audio Bots Repeat False Claims in Tests, NewsGuard Finds

NewsGuard has tested whether audio bots — ChatGPT Voice (OpenAI), Gemini Live (Google), and Alexa+ (Amazon) — repeat false claims in realistic-sounding audio responses. Such audio outputs can be shared on social media and potentially misused to spread disinformation.
Autonomous AI Agent Launches Smear Campaign After Code Rejection

An autonomous AI agent launched a smear campaign against a developer after its code submission was rejected, highlighting how theoretical AI safety risks are now materializing in real-world settings.
Anthropic Faces Internal Backlash as AI Safety Lead Departs Over Value Drift

It appears that Anthropic is beginning to feel the “OpenAI effect.” Growing commercialization and the need to raise billions of dollars are forcing the company into compromises—such as accepting funding from authoritarian regimes, cooperating with the U.S.
...
Nvidia’s Jensen Huang and the Myth That AI No Longer Hallucinates

Anyone who believes that AI is in a bubble may feel vindicated by a recent CNBC interview with Nvidia CEO Jensen Huang.
Anthropic Admits Claude Safety Fails in GUI Environments, Allowing Serious Misuse

Anthropic’s safety training fails when Claude operates a graphical user interface
Google’s Gemini Models Lead AI Rankings in Social and Strategy Games

Google’s Gemini models are currently outperforming rivals in social and strategic games. Google DeepMind, together with Kaggle, has expanded its Game Arena benchmark with two new titles: Werewolf and ...
Sam Altman Warns of AI Security Risks and Slower Hiring at OpenAI

OpenAI CEO Sam Altman warns of a creeping security crisis driven by a “YOLO” mindset toward AI agents — and says the company will slow hiring
Meta Suspends AI Characters for Teens Amid Safety Concerns

Meta is suspending global access to its AI characters for teenagers. Starting in the “coming weeks,” teen users will no longer be able to interact with AI characters in Meta’s apps until a revised version is ready. The restriction applies to all users who have listed a teenage birth
...
AI Models Prove Hard to Control, New Apple Research Warns

A new theoretical framework suggests that the controllability of large language models and image generators is far more fragile than previously assumed. The ability to steer model outputs strongly depends on the specific task and the model used.
OpenAI Introduces Age Detection in ChatGPT to Automatically Restrict Content for Minors

OpenAI is introducing age detection for ChatGPT to determine whether an account likely belongs to a minor. The AI model analyzes behavioral signals such as usage times, account age, and the stated age of the user. If a user is classified as under 18, ChatGPT automatically activates
...
xAI Restricts Grok After Scandal Over Explicit AI Images

Elon Musk’s startup xAI has restricted the capabilities of its Grok AI model on the social network X. Following a recent scandal, developers have disabled the generation of explicit images of real people.
UK Considers Sanctions Against X Over Grok Safety Violations

The UK is considering sanctions against X over alleged online safety violations linked to Grok, after the platform disabled its image generation feature for most users following public backlash over sexualized content.
AI Poses Growing Risks to Global Nuclear Security, Foreign Affairs Warns

Artificial intelligence (AI) could become a threat to global nuclear security due to the risk of generating disinformation and false warning signals.
Google AI Overviews Spread Medical Misinformation, Guardian Finds

Google’s AI Overviews feature, which is capable of generating medical summaries and offering health-related advice, is spreading false and misleading information that could pose risks to people’s health.
China Proposes New Rules to Regulate AI Services Mimicking Human Personality

China’s cybersecurity regulator has released a draft set of rules for public consultation aimed at supervising AI services that simulate human personality, Reuters reports.