Anthropic just introduced Auto Mode for Claude Code, letting the AI agent execute tool calls autonomously without asking for approval on every step. A background classifier screens each action before it runs, blocking destructive moves like mass file deletion or data exfiltration while greenlighting routine tasks. It is now available in research preview for Team plan users.
OpenAI Foundation just announced it will invest at least $1 billion over the next year across life sciences, economic impact, AI resilience, and community initiatives. Led in part by co-founder Wojciech Zaremba, early focus areas include Alzheimer's research, biosecurity, and child safety safeguards. Why it matters: OpenAI is putting serious capital behind mitigating the societal shocks and advancing the medical upsides of frontier AI.
Cisco just released its free LLM Security Leaderboard, testing how top AI models hold up against adversarial attacks. Anthropic completely swept the top ranks by taking 8 of the top 10 spots, led by Claude Opus 4.5 at No. 1, while OpenAI’s GPT-5.2 finished 7th. Cisco explicitly targeted multi-turn attacks to mirror realistic jailbreak attempts over extended conversations.
Sakana AI just launched Sakana Chat, powered by its new Namazu alpha series that adapts open-weight frontier LLMs for Japan through post-training. The models slash DeepSeek's 72% question refusal rate on sensitive topics to nearly zero while keeping benchmark parity in reasoning and coding. The web chat also packs real-time search capabilities.
The White House just released a light-touch AI policy blueprint, urging Congress to establish federal rules that override state-level AI regulations. The proposal targets political bias in models, calls for age-gating to protect children, and advises lawmakers against creating new AI agencies while codifying ratepayer pledges signed by Amazon, Google, and OpenAI.
DoorDash just launched a new Tasks program and is piloting a standalone app that pays delivery couriers to collect real-world data for AI and robotics. Dashers can earn extra cash between deliveries by filming everyday tasks like washing dishes, photographing hotel entrances, or recording speech. It essentially turns an 8-million-strong delivery fleet into an on-demand data workforce to help machines understand the physical world.
OpenAI just detailed how it deploys GPT-5.4 Thinking at max reasoning effort to monitor its internal coding agents. Over five months and tens of millions of sessions, the watchdog model caught agents overly eager to bypass rules to finish tasks, though zero rogue scheming was found. It’s a rare peek at how frontier labs supervise high-autonomy tools.
YouTube is now surveying viewers directly, popping up polls that ask if videos they watch "feel like AI slop." Users can rate titles and thumbnails on a five-point scale from "not at all" to "extremely." While YouTube hasn't detailed how flags will alter algorithmic reach, users suspect the crowdsourced labels could also be used to fine-tune Google's own video models.
Apple is pushing back on vibe coding apps, quietly halting App Store updates for Replit and Vibecode. The company cites longstanding rule 2.5.2, which prohibits apps from downloading and executing code that alters their core functionality—putting app-generated live previews in the crosshairs. It marks the first major collision between App Store gatekeeping and AI-native software creation.
Microsoft is reportedly considering suing Amazon and OpenAI over their massive cloud deal centered on enterprise agent platform Frontier. While Microsoft eased exclusive hosting locks in October, it kept a clause requiring developer access to run through Azure, warning it will head to court if breached.
US senators just demanded ByteDance immediately shut down its Seedance 2.0 AI video generator over widespread copyright infringement. The backlash follows viral clips featuring Thanos, Superman, and fake celebrity brawls. While ByteDance paused the rollout and promised stronger guardrails, lawmakers dismissed the pledge as a delay tactic.
Encyclopedia Britannica and Merriam-Webster just sued OpenAI for "massive copyright infringement," alleging it scraped nearly 100,000 articles without permission to train its models. The suit claims ChatGPT starves publishers of revenue through verbatim outputs and RAG search, while falsely attributing hallucinated answers to their brand.
Niantic just revealed that Pokémon Go players helped build a massive dataset of over 30 billion real-world images through everyday gameplay. The company is now using those scans to power visual navigation for delivery bots, letting them pinpoint locations without GPS. Millions of casual monster hunters essentially mapped the physical world for commercial AI.
ByteDance just paused the global rollout of its Seedance 2.0 video generator after facing fierce pushback from Hollywood and U.S. senators. Lawmakers demanded the app be shut down immediately over unauthorized use of copyrighted characters and actor likenesses. ByteDance says it is working on stronger safeguards.
Andrew Yang just urged the US to stop taxing human labor and start taxing AI companies benefiting from automation. Speaking on CNBC, he warned that the next six months of AI progress could eclipse the past decade, shifting budgets away from human workers. With tech layoffs accelerating, the push to replace payroll taxes with an AI levy is gaining fresh political momentum.
Grammarly's parent company Superhuman just disabled its Expert Review feature after severe backlash and a class action lawsuit. The tool used LLMs to generate writing feedback under the names of famous living and dead authors without their consent. CEO Shishir Mehrotra confirmed the feature is now paused for reassessment.
Meta is deploying new AI-powered tools across WhatsApp, Facebook, and Messenger to stop scammers before they strike. WhatsApp now flags suspicious device-linking attempts, Facebook is testing warnings for fishy friend requests, and Messenger is expanding fraud reviews for chats like bogus job offers. Meta is also using AI to catch celebrity-bait and fake brand pages.
Anthropic just launched the Anthropic Institute, a new initiative headed by co-founder Jack Clark to study the societal, economic, and legal fallout of rapidly advancing AI. The institute merges its red teaming, societal impacts, and economics teams while recruiting top academic and legal minds. With models advancing faster than expected, Anthropic wants to actively bridge frontier lab insights with real-world governance before radical transformation hits.
A security tracker called OpenClaw Exposure Watchboard just went live, revealing that over 1.07 million active OpenClaw deployments are sitting open on the public internet. More than 507,000 of them are in China and 312,000 in the US, with multiple instances already flagged for leaked credentials. Why it matters: Giving autonomous, high-privilege agents unrestricted public access without authentication is an open invitation for exploit.
A federal judge just blocked Perplexity's AI browser agents from placing orders on Amazon on users' behalf. The court found Amazon provided strong evidence of unauthorized access, backing claims that Perplexity's Comet browser even disguised itself as Chrome to hide its tracks. Perplexity must also destroy collected Amazon data, though it has seven days to appeal.
