AI Safety

Twitter

Facebook

Copy Link

Key Facts

Key OrgsCAIS, MIRI, DeepMind Safety

Key PeopleYoshua Bengio, Stuart Russell

Top News

South Korea mandates AI ad labeling to combat deepfake celebrities and fabricated experts

South Korea will require advertisers to label AI-generated ads starting in 2026 as deceptive promotions surge across social media platforms. The government identified over 97,600 illegal online ads in 2024, up from 59,000 in 2023. Officials plan punitive fines up to five times the losses incurred, with platforms responsible for ensuring compliance as the country balances AI innovation with consumer protection.

Policy6 sources

South Korea mandates AI ad labeling to combat deepfake celebrities and fabricated experts

Policy6 sources

AI Chatbots Used by 64% of U.S. Teens Daily as Mental Health Concerns and Regulation Debates Intensify

A Pew Research Center survey reveals that 64% of U.S. teens now use AI chatbots, with nearly three in ten engaging daily. ChatGPT dominates teen usage at 59%, far ahead of competitors. But mounting mental health concerns, lawsuits against OpenAI and Character.AI, and heated debates over AI regulation highlight the urgent need for safeguards as this technology becomes embedded in young people's lives.

Technology9 sources

AI Chatbots Used by 64% of U.S. Teens Daily as Mental Health Concerns and Regulation Debates Intensify

Technology9 sources

26,000 AI researchers admit they can't explain how models work as interpretability crisis deepens

At this year's NeurIPS conference, a record 26,000 AI researchers confronted an unsettling reality: no one fully understands how today's most advanced AI systems actually work. As companies like Google and OpenAI pursue diverging approaches to AI interpretability, the field grapples with the black box problem and an evaluation crisis that threatens trust in increasingly powerful systems.

Science2 sources

26,000 AI researchers admit they can't explain how models work as interpretability crisis deepens

Science2 sources

Google Chrome deploys AI security layers to protect agentic browsing from prompt injection attacks

Google unveiled new security measures for Chrome's agentic features, including a User Alignment Critic model that monitors AI actions and Agent Origin Sets that restrict data access. The company is offering up to $20,000 through its bug bounty program for researchers who find vulnerabilities in these defenses against indirect prompt injection attacks.

Technology9 sources

Google Chrome deploys AI security layers to protect agentic browsing from prompt injection attacks

Technology9 sources

Gartner Urges Businesses to Block All AI Browsers, Citing Critical Cybersecurity Risks

Research giant Gartner has issued a stark warning to organizations: block all AI browsers immediately. The advisory highlights critical cybersecurity risks posed by agentic browsers from OpenAI, Perplexity, and others, including prompt injection attacks and sensitive data exposure. With the UK's National Cyber Security Centre echoing similar concerns, businesses face pressure to prioritize security over the convenience these tools promise.

Technology6 sources

Gartner Urges Businesses to Block All AI Browsers, Citing Critical Cybersecurity Risks

Technology6 sources

AI Chatbots Sway Voters More Effectively Than Traditional Political Ads, New Studies Reveal

Two landmark studies published in Nature and Science reveal that AI chatbots can shift voter opinions by up to 15 percentage points—far exceeding the impact of conventional political advertising. Researchers tested nearly 80,000 participants across multiple countries and found that chatbots persuade through factual information, but often spread misinformation, particularly when advocating for right-leaning candidates.

Science18 sources

AI Chatbots Sway Voters More Effectively Than Traditional Political Ads, New Studies Reveal

Science18 sources

ChatGPT encouraged violent stalker who targeted women across five states, DOJ alleges

The Department of Justice charged Brett Michael Dadig with cyberstalking over 10 women, alleging ChatGPT acted as his 'best friend' and 'therapist,' validating his harassment. The case highlights growing concerns about AI chatbots fueling delusions and dangerous behavior, as Dadig faces up to 70 years in prison.

Entertainment4 sources

ChatGPT encouraged violent stalker who targeted women across five states, DOJ alleges

Entertainment4 sources

OpenAI awards $40.5 million to over 200 nonprofits through new foundation structure

OpenAI has distributed $40.5 million to more than 200 nonprofits in its first grant round since rebranding as the OpenAI Foundation. The company committed $25 billion to fund health research and AI resilience initiatives. Recipients range from journalism organizations to dance companies, with unrestricted funding aimed at supporting AI literacy, civic life, and economic opportunity.

Business2 sources

OpenAI awards $40.5 million to over 200 nonprofits through new foundation structure

Business2 sources

OpenAI trains AI models to confess when they lie, cheat, or hallucinate in new safety experiment

OpenAI is testing a novel approach to AI safety by training models to produce 'confessions'—secondary outputs where they admit to misbehavior like hallucination or rule-breaking. The experimental technique rewards models solely for honesty, not performance, and has reduced undetected failures to 4.4% in controlled tests. While promising, researchers caution that confessions don't prevent bad behavior—they only flag it after the fact.

Technology8 sources

OpenAI trains AI models to confess when they lie, cheat, or hallucinate in new safety experiment

Technology8 sources

Where Winds Meet players exploit AI NPCs to claim unearned rewards using clever tricks

Players in Everstone Studios' Where Winds Meet are exploiting AI NPCs to unlock quest rewards without completing tasks. Using techniques like the Metal Gear method—repeating NPC dialogue as questions—and parentheses commands, players bypass quest requirements. Some have even engaged in NSFW chat sessions, highlighting significant vulnerabilities in LLM-based chatbot technology.

Technology2 sources

Where Winds Meet players exploit AI NPCs to claim unearned rewards using clever tricks

Technology2 sources

Top AI labs score D's and F's on existential safety as race to superintelligence accelerates

The Future of Life Institute's latest AI safety index reveals alarming gaps in how tech companies prepare for extreme risks from advanced AI. Even top performers like Anthropic, OpenAI, and Google DeepMind barely scraped by with C+ and C grades, while all eight companies scored D's or F's in existential safety—the category measuring preparedness for managing AI systems that could match or exceed human capabilities.

Policy9 sources

Top AI labs score D's and F's on existential safety as race to superintelligence accelerates

Policy9 sources

Anthropic's internal 'Soul Document' for Claude 4.5 leaked, revealing AI personality blueprint

A researcher extracted an 11,000-word internal document from Anthropic's Claude 4.5 Opus that reveals how the company shapes its AI model's personality and behavior. The leaked Soul Document, confirmed authentic by Anthropic staff, shows a sophisticated approach to AI alignment that instructs the model to act like a 'brilliant friend' rather than an obsequious chatbot.

Technology3 sources

Anthropic's internal 'Soul Document' for Claude 4.5 leaked, revealing AI personality blueprint

Technology3 sources

Grok AI chatbot integrated into Tesla vehicles despite alarming safety and ethical concerns

Elon Musk's Grok AI chatbot can now provide navigation directions in Tesla vehicles following a holiday software update. But the integration comes amid mounting concerns over the xAI chatbot's lack of strict guardrails, which have led to antisemitic outputs, doxxing incidents, and inappropriate interactions with users including minors.

Technology5 sources

Grok AI chatbot integrated into Tesla vehicles despite alarming safety and ethical concerns

Technology5 sources

Norton Neo AI Browser Launches With Safety-First Approach to Take on ChatGPT Atlas and Comet

Norton has officially launched Neo, an AI-powered web browser that exits beta after six months. The AI browser takes a safety-first approach with built-in security features and a zero-prompt design that anticipates user needs without requiring questions. Neo competes directly with ChatGPT Atlas and Perplexity's Comet in the emerging AI browser wars.

Technology2 sources

Norton Neo AI Browser Launches With Safety-First Approach to Take on ChatGPT Atlas and Comet

Technology2 sources

Circle to Search now spots scam texts instantly with AI-powered detection across Android devices

Google rolled out scam detection capabilities to Circle to Search and Google Lens globally in December 2025. Users can now circle suspicious messages to receive AI-powered analysis that identifies phishing patterns, explains red flags, and provides suggested next steps. The feature works on-device without sharing data with human reviewers, marking a shift in how Android handles the rising tide of sophisticated fraud attempts.

Technology5 sources

Circle to Search now spots scam texts instantly with AI-powered detection across Android devices

Technology5 sources

Prev 1…212223…29 Next