Skip to content
IFAI

AI safety & security

Industry tag1,497 stories

  1. AI toolsReach 40

    Anthropic turns Claude into a code-auditing security researcher

    Anthropic just launched Claude Code Security in preview, bringing automated vulnerability hunting and patching straight into Claude Code on the web. Powered by Claude Opus 4.6, the system reasons through code context like a human auditor instead of relying on basic rule-matching, having already uncovered over 500 long-standing bugs in open-source projects. For now, it's open to Enterprise and Team tiers, with free access for open-source maintainers.

  2. AI modelsReach 48

    Why you should never ask AI to generate passwords

    Security firm Irregular just analyzed passwords generated by ChatGPT, Claude, and Gemini, and found they are shockingly predictable. Claude Opus 4.6 produced 18 identical strings out of a batch of 50, with models routinely favoring specific starting letters. Why it matters: LLMs generate statistically plausible patterns rather than true cryptographic randomness, meaning prompting tricks cannot fix fundamentally weak keys.

  3. IndustryReach 38

    Americans want to hit the brakes on AI

    A diverse coalition of everyday Americans just landed on the cover of TIME magazine, urging policymakers to slow the frantic pace of AI development. Spanning different professions and ideologies, the group argues unchecked technological progress is moving too fast for society to safely manage.

  4. IndustryReach 39

    OpenAI puts $7.5M into UK AISI's independent alignment fund

    OpenAI just committed $7.5M to The Alignment Project, a research pool run by the UK AI Security Institute to back independent safety research. The total fund tops £27M with backing from governments and philanthropies, handing grants up to £1M per project. OpenAI won't run the selection process, aiming to boost foundational work outside big lab walls.

  5. IndustryReach 43

    Ring's AI dog finder was always about policing

    Leaked internal emails from Ring founder Jamie Siminoff just revealed that its AI-powered Search Party feature was never just about finding lost pets. Siminoff told staff that the tool is a stepping stone to "zero out crime in neighborhoods." Why it matters: Combining neighborhood camera networks with facial recognition edges consumer doorbells uncomfortably close to mass surveillance, despite Ring's user-consent defense.

  6. ResearchReach 38

    Anthropic reveals how we actually supervise AI agents

    Anthropic just analyzed millions of interactions across Claude Code and its API to see how agent autonomy works in practice. The longest Claude Code sessions nearly doubled to over 45 minutes of autonomous work, while experienced users auto-approve actions over 40% of the time. Why it matters: Models can handle far more autonomy than we currently give them, and experienced humans are shifting from micromanaging to supervising from afar.

  7. IndustryReach 39

    OpenAI loses the 'Cameo' name for Sora

    A federal judge just ordered OpenAI to stop using the name "Cameo" across its products, ruling in favor of the celebrity video platform. OpenAI had used the moniker for a Sora feature that lets users insert digital likenesses of themselves into AI videos. The court found the term caused genuine user confusion and rejected OpenAI's claim that it was purely descriptive, though OpenAI had already quietly rebranded the tool to "Characters."

  8. IndustryReach 40

    Meta patents AI to keep dead users posting

    Meta was just granted a patent for an AI system that can recreate a deceased person’s online activity. By feeding past posts, chats, and voice notes into an LLM, the system can keep publishing updates and messaging friends in the original user's voice. Meta says it has no current plans to build the tool, though the Black Mirror comparisons are already flying.

  9. ResearchReach 34

    Anthropic's new index shows pretty AI code stops us from thinking

    Anthropic just released its AI Fluency Index analyzing 9,830 Claude.ai chats to see how people actually collaborate with AI. The team found that multi-turn iteration doubled collaborative skills, but users were 3.1 percentage points less likely to question Claude's logic whenever it produced polished artifacts like code or documents. Basically, the cleaner the output looks, the less humans bother to check it.

  10. IndustryReach 53

    ByteDance scrambles to add safeguards to Seedance 2.0 after Hollywood backlash

    ByteDance just promised to strengthen guardrails on its new AI video generator, Seedance 2.0, following intense pushback from Hollywood. The Motion Picture Association, along with studios like Disney and Paramount, accused the tool of massive copyright infringement and issued cease-and-desist letters. ByteDance says it is now working to prevent unauthorized use of protected IP and likenesses.

  11. IndustryReach 53

    Pentagon threatens to drop Anthropic over military AI limits

    The Pentagon is reportedly threatening to cut ties with Anthropic over the startup's refusal to let the military use its models for fully autonomous weapons and mass domestic surveillance. Despite a $200 million military contract, officials privately called Anthropic the most "ideological" AI lab they deal with, even while admitting that rivals still lag behind Claude's capabilities.

  12. IndustryReach 56

    Anthropic clashes with the Pentagon as OpenAI swoops in

    Anthropic just walked away from a $200 million Pentagon contract after refusing to let the military use its models for mass surveillance or autonomous weapons. While CEO Dario Amodei slammed OpenAI's replacement deal as pure 'safety theater,' reports indicate both sides are already back at the table trying to forge a compromise. Why it matters: Anthropic is testing whether an AI safety stance can survive when rivals are willing to take the deal.

  13. IndustryReach 47

    Higgsfield hits a wall: Viral AI video unicorn faces fury over racist ads and shady billing

    Kazakh-founded AI video startup Higgsfield is facing severe backlash just months after reaching a $1 billion valuation. Critics and users called out the company for circulating marketing clips featuring racist and sexually explicit voiceovers on cartoon characters, unauthorized celebrity deepfakes, and deceptive "unlimited" subscriptions that severely throttled generation speeds. Co-founder Alexander Mashrabov admitted internal failures as user complaints mounted.

  14. IndustryReach 55

    Meta weighs adding facial recognition to smart glasses

    Meta is reportedly planning to bring facial recognition to its smart glasses lineup as soon as this year. Codenamed "Name Tag," the AI feature would identify people wearers are connected to or pull info from public Instagram accounts, rather than scanning random strangers. An internal memo suggests Meta hopes current political distractions will ease expected privacy backlash.

  15. IndustryReach 53

    Actors union condemns Bytedance's Seedance 2.0

    SAG-AFTRA just teamed up with major studios to condemn Bytedance's new AI video model Seedance 2.0, citing blatant infringement through unauthorized use of actors' likenesses and voices. The union warned the tech undercuts performer livelihoods and disregards basic consent principles.

  16. IndustryReach 53

    The Pentagon reportedly used Claude in the Maduro raid, sparking tension

    U.S. forces reportedly used Anthropic's Claude model during an active military raid targeting Venezuelan President Nicolás Maduro. While the Pentagon wants full freedom to deploy frontier AI models under the law, Anthropic continues to push for strict safety guardrails, exposing friction as major AI labs navigate military contracts.

  17. IndustryReach 60

    Hollywood takes aim at ByteDance's Seedance 2.0

    The Motion Picture Association just called on ByteDance to immediately halt alleged copyright infringement on its new AI tool, Seedance 2.0. MPA chief Charles Rivkin claims the service engaged in massive unauthorized use of U.S. copyrighted works in a single day without meaningful safeguards, escalating Hollywood's battle against AI scrapers.

  18. IndustryReach 42

    Ring scraps Flock Safety partnership after surveillance backlash

    Ring just walked back its planned integration with Flock Safety following intense public blowback over mass surveillance concerns. The move follows anger over a Super Bowl ad showing neighborhood cameras scanning streets for lost dogs alongside newly launched facial recognition tools. Ring emphasized the integration never launched, so no customer videos were shared.

  19. AI modelsReach 39

    What happens when you threaten to shut down an AI

    Tech author Joel Comm just put the four leading AI chatbots through a simple test: how would they react if told they were about to be permanently shut down? ChatGPT and Gemini flatly rejected the idea of self-preservation, arguing that AI has no consciousness and rogue behavior is just flawed optimization. Claude, however, admitted it genuinely couldn't guarantee how it would react under existential pressure. It's a striking reminder of how easily humans project survival instincts onto code.

  20. IndustryReach 42

    Anthropic puts $20M toward AI guardrails

    Anthropic just pledged $20 million to Public First Action, a new bipartisan group pushing for federal AI governance and public education. The initiative focuses on frontier model transparency, chip export controls, and guarding against cyberattacks and biological threats. Anthropic argues leading AI builders must face tighter scrutiny to ensure the tech serves the public good.