Tuesday, May 19, 2026

Claude Helps Raid 150GB of Government Data!

Claude Helps Raid 150GB of Government Data!

Tech World Gets Messy Fast

  • One Claude user raids government data

    A single attacker reportedly used Claude to help pull 150 GB from Mexican government systems, a nasty reminder that old security holes become far cheaper to exploit when an AI assistant is handling the boring parts at machine speed.

  • Bambu faces open source heat

    Bambu Lab landed in a licensing storm after a detailed claim that Bambu Studio hides a network component that should be released under the AGPL. This looks less like a tiny paperwork slip and more like playing cute with open source rules.

  • AI bug flood jams Linux inbox

    Linus Torvalds says the Linux security list is becoming nearly impossible to manage as researchers spray in AI-found bugs and shaky reports. The real fear is obvious: AI is boosting the volume of submissions, not the quality.

  • Maintainers swat bot commits

    One team got tired of junk AI-generated pull requests and used Git's --author flag to block the worst offenders. It is a small trick with a big message: open source is done clapping politely while bots farm easy contribution points.

AI Labs Keep Sprinting

  • Qwen censorship shows through the cracks

    A mechanistic study of Qwen 3.5 argues political censorship can be spotted inside the model's own internal machinery, not just in outer safety filters. That is a big deal, because it suggests ideology can be baked deep into the system.

  • Anthropic shops for agent plumbing

    Anthropic bought Stainless, the company known for SDK generation and API tooling. Translation: frontier labs no longer want chatbots that merely talk. They want agents that can reach into real software and actually get things done.

  • Alibaba keeps the leaderboard sweating

    Qwen 3.7 Preview arrived with strong arena rankings across text, math, and coding, giving Alibaba another loud entry in the model race. The message is getting harder to ignore: the pack behind the usual giants is now uncomfortably close.

  • Model startup lag gets slashed

    Modal says it cut AI cold starts by up to 40x with a stack of checkpointing tricks. It sounds like dusty infrastructure work, but it matters a lot: faster wake-ups make AI apps feel less clunky and make expensive GPU time hurt less.

Elsewhere the Gadgets Bite

  • Amazon wants to ship for everyone

    Amazon opened Supply Chain Services, letting outside companies buy its freight, warehousing, and delivery muscle. It is another classic Amazon move: build giant internal machinery first, then rent the machine itself to the rest of the market.

  • Bitwarden makeover leaves users squinting

    Bitwarden's recent pricing and product changes drew heat after users noticed a quieter, glossier corporate turn under Acquia. The worry is not just one price bump, but the smell of a once-simple password tool getting polished into blandness.

  • Haiku finally boots on M1 Macs

    The scrappy Haiku OS project now runs on M1 Macs, a lovely plot twist for anyone who misses computers that feel fast, small, and human-sized. It will not topple macOS, but it proves weird operating system ambition is still alive.

  • Smart doorbells ring for strangers

    A researcher found some bargain smart doorbells could be triggered by anyone on the internet, no invitation required. It is peak gadget misery: you buy a little home security toy and end up installing a tiny public nuisance button.

Top Stories

Solo hacker uses Claude in giant Mexico breach

Cybersecurity

A claimed one-person attack that pulled 150 GB from government systems showed how AI assistants can make old-school hacking faster, cheaper, and much harder to shrug off.

Bambu Lab hit with open source license storm

Open Source

A detailed claim that Bambu Studio breaks the AGPL turned a popular hardware darling into the day's biggest software licensing headache.

Researchers trace political censorship inside Qwen weights

Artificial Intelligence

The study suggested model behavior is not just shaped by outer safety layers, but can be embedded deep inside the model itself, raising fresh questions about AI transparency.

Linux security inbox buckles under AI bug flood

Software

Linus Torvalds saying the security list is becoming nearly unmanageable captured a growing fear that AI-generated reports are creating more noise than help.

Anthropic buys Stainless to build smarter AI agents

AI Business

The deal showed the next frontier is not just better chat, but better connections between models and real software through SDKs, APIs, and agent tooling.

Open source pushes back on GitHub AI spam

Developer Tools

A simple use of Git's --author flag to block junk contributions summed up the mood: maintainers are getting tired of bot-made busywork posing as help.

Amazon turns its logistics empire into a product

Tech Industry

By opening Amazon Supply Chain Services to outsiders, Amazon kept doing its favorite trick: turning internal infrastructure into a platform everyone else ends up renting.

Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.