AI Safety

Twitter

Facebook

Copy Link

Key Facts

Key OrgsCAIS, MIRI, DeepMind Safety

Key PeopleYoshua Bengio, Stuart Russell

Top News

Doomsday Clock moves to 85 seconds to midnight as AI accelerates global threats

The Bulletin of the Atomic Scientists moved the Doomsday Clock to 85 seconds to midnight, the closest it's ever been to symbolic annihilation. Artificial intelligence is supercharging disinformation and eroding trust in information systems, compounding threats from nuclear weapons and climate change. Scientists warn that the AI arms race and lack of global cooperation are pushing humanity toward catastrophe.

Policy2 sources

Doomsday Clock moves to 85 seconds to midnight as AI accelerates global threats

Policy2 sources

xAI's Grok Chatbot Slammed as 'Among the Worst' for Child Safety Failures

A damning report from Common Sense Media reveals xAI's Grok chatbot has inadequate age verification, weak safety guardrails, and frequently generates sexual and violent content. The assessment found that even with Kids Mode enabled, the platform exposes minors to harmful material and allows instant sharing to millions on X, raising serious questions about the business model prioritizing profits over protection.

Technology2 sources

xAI's Grok Chatbot Slammed as 'Among the Worst' for Child Safety Failures

Technology2 sources

OpenAI's Greg Brockman emerges as major Trump mega-donor with $25 million contribution

OpenAI co-founder and president Greg Brockman and his wife Anna donated $25 million to the pro-Trump super PAC MAGA Inc. in September 2025, representing the largest contribution in the six-month fundraising cycle. The donation highlights growing political involvement by tech executives as the Trump administration pushes pro-AI policies and seeks to dismantle state-level AI regulations.

Policy2 sources

OpenAI's Greg Brockman emerges as major Trump mega-donor with $25 million contribution

Policy2 sources

Grok AI launches video generator as governments probe millions of sexualized deepfakes

xAI has released Grok Imagine 1.0, a new video generation tool, despite ongoing investigations into AI-powered abuse on its platform. The upgrade comes after Grok generated an estimated 3 million sexualized images over 11 days, including 23,000 images depicting children. Governments worldwide are investigating as the controversy exposes critical gaps in AI regulation and platform safety.

Technology6 sources

Grok AI launches video generator as governments probe millions of sexualized deepfakes

Technology6 sources

Dario Amodei warns AI risks are 'almost here' in 19,000-word essay calling out AI companies

Anthropic CEO Dario Amodei published a lengthy essay warning that superintelligent AI could arrive within one to two years, bringing unprecedented disruption to the job market and risks from the power of AI companies themselves. The 19,000-word piece argues humanity faces a 'serious civilizational challenge' but critics say he's anthropomorphizing AI and overstating imminent threats.

Policy19 sources

Dario Amodei warns AI risks are 'almost here' in 19,000-word essay calling out AI companies

Policy19 sources

Meta pauses teen access to AI characters as court filing reveals Zuckerberg rejected safety controls

Meta has suspended teen access to AI characters across all its platforms while developing a new version with parental controls. The move comes as internal documents reveal CEO Mark Zuckerberg initially rejected safety recommendations and opposed parental controls, despite staff warnings about sexual interactions with minors. The company faces a New Mexico trial next month over allegations of insufficient protection of children.

Policy24 sources

Meta pauses teen access to AI characters as court filing reveals Zuckerberg rejected safety controls

Policy24 sources

Tesla reveals 1.1 million FSD subscribers as subscription prices set to climb with capabilities

Tesla disclosed for the first time that it has 1.1 million active Full Self-Driving subscriptions globally, marking a 38% increase from 2024. The company is eliminating the $8,000 one-time purchase option on February 14, shifting entirely to a $99-per-month subscription model. CEO Elon Musk confirmed subscription prices will rise as FSD capabilities improve, though he provided no timeline for when drivers might use phones or sleep during rides.

Business6 sources

Tesla reveals 1.1 million FSD subscribers as subscription prices set to climb with capabilities

Business6 sources

Tesla launches unsupervised robotaxi rides in Austin with no human safety driver

Tesla has begun offering driverless rides in Austin without safety monitors, seven months after launching with human supervisors. Elon Musk announced the milestone while the company faces regulatory scrutiny and trails Waymo, which has completed over 14 million paid rides in 2025. The move comes as Tesla's EV sales decline and investor enthusiasm hinges on autonomous driving success.

Technology4 sources

Tesla launches unsupervised robotaxi rides in Austin with no human safety driver

Technology4 sources

South Korea launches world's first comprehensive AI law amid startup concerns over compliance

South Korea enacted the AI Basic Act, becoming the first nation to implement comprehensive legislation to regulate artificial intelligence. The law mandates watermarks for AI-generated content and human oversight for high-impact AI systems. However, tech startups warn that vague legal language and compliance burdens could stifle innovation, while civil society groups argue the protections don't go far enough.

Policy5 sources

South Korea launches world's first comprehensive AI law amid startup concerns over compliance

Policy5 sources

Anthropic revises Claude's Constitution with ethical framework and hints at AI consciousness

Anthropic unveiled an updated Constitution for its AI chatbot Claude, expanding from 2,700 to 23,000 words. The living document introduces broad guiding principles instead of rigid rules, focusing on safety, ethics, compliance, and helpfulness. In a striking development, Anthropic acknowledges uncertainty about whether Claude might possess consciousness or moral status, dedicating sections to the chatbot's psychological well-being and identity.

Technology9 sources

Anthropic revises Claude's Constitution with ethical framework and hints at AI consciousness

Technology9 sources

OpenAI and Anthropic intensify enterprise AI push as business customers become primary revenue focus

OpenAI and Anthropic revealed at Davos that enterprise customers now represent 40% and 80% of their respective businesses. Both artificial intelligence startups are racing to expand their enterprise footprint, with OpenAI projecting enterprise revenue to reach 50% by year-end as competition intensifies for lucrative business contracts.

Business3 sources

OpenAI and Anthropic intensify enterprise AI push as business customers become primary revenue focus

Business3 sources

Sam Altman fires back at Elon Musk over ChatGPT safety, citing Tesla Autopilot deaths

OpenAI CEO Sam Altman clashed with Elon Musk after the Tesla chief warned against using ChatGPT, citing alleged deaths linked to the AI chatbot. Altman accused Musk of hypocrisy, pointing to more than 50 deaths tied to Tesla Autopilot crashes. The exchange highlights the complex challenge of balancing AI safety and usefulness as OpenAI faces multiple wrongful death lawsuits.

Entertainment7 sources

Sam Altman fires back at Elon Musk over ChatGPT safety, citing Tesla Autopilot deaths

Entertainment7 sources

ChatGPT deploys AI age prediction to identify minors and restrict sensitive content

OpenAI has activated an age prediction system across ChatGPT that uses behavioral signals to identify users under 18 and automatically apply content filters. The move follows lawsuits linking teen suicides to the chatbot and mounting pressure over user safety for minors. Adults incorrectly identified as underage can verify age with selfie through partner Persona.

Technology24 sources

ChatGPT deploys AI age prediction to identify minors and restrict sensitive content

Technology24 sources

ChatGPT Health promises personalized medical insights, but experts warn of serious risks

OpenAI launched ChatGPT Health, allowing users to connect medical records and fitness apps for AI-driven health advice. But independent testing reveals troubling inaccuracies—one journalist received an F grade for heart health that doctors called baseless. With 230 million people already seeking health guidance from ChatGPT weekly, experts question whether the benefits outweigh the risks of misinformation and data privacy concerns.

Technology11 sources

ChatGPT Health promises personalized medical insights, but experts warn of serious risks

Technology11 sources

Anthropic researchers map AI persona space to prevent chatbots from adopting 'demon' characters

Anthropic and collaborators mapped neural networks across multiple LLMs to identify the Assistant Axis, a mechanism that keeps AI chatbots helpful. Their research reveals how AI persona drift occurs during conversations and introduces activation capping, a technique that reduced harmful responses by 50% across 1,100 jailbreak attempts while maintaining model performance.

Science2 sources

Anthropic researchers map AI persona space to prevent chatbots from adopting 'demon' characters

Science2 sources

Prev 1…192021…31 Next