Teen Hackers Expose AI Secrets, Physicists Chase Elusive Glueballs

A group of teenagers pulled off one of the most consequential supply-chain attacks in recent memory — and they did it in 40 minutes.

The crew, known as TeamPCP, managed to poison LiteLLM, a widely used open source tool that helps developers build AI-powered software. During a narrow window in March, anyone who downloaded LiteLLM from the official Python Package Index repository got a compromised version that silently scraped their machine's memory and shipped the contents off to attacker-controlled servers. The result: credentials belonging to more than 2,500 organizations, including Microsoft, Amazon, Cisco, Samsung, and Salesforce, all sitting in a 195TB data dump.

Security firms CloudSEK and Hudson Rock published their findings this week after independently analyzing the leaked data. What they found was staggering — cloud keys, SSH keys, Kubernetes secrets, repository tokens, package publishing credentials, and AI provider keys, all ripe for exploitation. Roughly 434,000 CI/CD pipelines had credentials exposed during that brief infection window.

LiteLLM wasn't the original target. TeamPCP first compromised Trivy, a popular open source vulnerability scanner with a somewhat ironic job description. From there, the infection spread to KICS and the Telnyx Python SDK, creating a cascading supply-chain failure that hit organizations who had no idea they were running poisoned software.

Independent security researcher Kevin Beaumont confirmed the legitimacy of the data after reviewing it himself. His take was blunt: this wasn't AI being the threat — it was organizations so fixated on shipping AI products fast that they let basic DevOps security rot. The teenagers didn't need sophisticated zero-days. They needed patience and a target-rich environment of companies cutting corners.

That framing matters. The AI development ecosystem has exploded in the past two years, pulling in enormous amounts of sensitive infrastructure under its umbrella. Tools like LiteLLM sit at the center of that ecosystem, trusted implicitly by engineering teams racing to build the next AI feature. That trust, apparently, is exactly what makes them attractive targets.

Tracking which organizations were actually hit proved harder than expected. Credentials in the dump were often tied to generic pipeline configurations rather than clearly labeled company assets. One email domain initially flagged as belonging to SiriusXM turned out to trace back to AdsWizz, a subsidiary — a reminder that modern software infrastructure is deeply nested and attribution is rarely clean.

The broader lesson here is one the security community has been shouting for years: the software supply chain is only as strong as its least-scrutinized dependency. A tool downloaded millions of times from an official repository is not inherently safe. And when the organizations using that tool are simultaneously handling the most sensitive credentials in their entire stack — AI API keys, cloud access tokens, deployment secrets — the blast radius of a single compromise becomes almost incomprehensible.

Teenagers, a 40-minute window, and 195 terabytes. The math should make every engineering team uncomfortable.

Most of the mass in your body doesn't actually come from matter. It comes from glue. Specifically, from the energy generated by gluons — the particles that hold quarks together inside protons and neutrons. Thanks to E=mc2, all that binding energy converts into mass, which means gluons are quietly responsible for most of what makes you physically real.

Now physicists think they may have finally spotted a particle made entirely of gluons, with no quarks involved at all. It's called a glueball, it's been predicted by quantum theory for decades, and researchers at the Beijing Spectrometer III experiment — BES III, for short — have just published what they're calling the strongest evidence yet that these things actually exist.

The findings appeared in a preprint on arXiv and were presented last week at the International Conference on High Energy Physics. The physics community has been cautiously excited, which in particle physics terms is roughly equivalent to everyone else losing their minds.

Glueballs are a direct prediction of quantum chromodynamics, the theory governing the strong nuclear force. QCD says gluons don't just bind to quarks — they also interact with each other. In principle, that means you could assemble a particle out of pure gluons, no quark required. Theorists have mapped out what such a particle should look like: zero spin, no electric charge, odd parity. Finding one that checks all those boxes in a real detector has been the challenge.

The BES III collider was built specifically to hunt for these signatures by studying the decay of J/psi particles, a type of meson first discovered in 1974. When J/psi particles decay, they spray out gluons and composite particles called hadrons — exactly the kind of messy, gluon-rich environment where a glueball might briefly flicker into detectable existence. Since the updated collider came online in 2008, it has recorded well over 226 million particle collision events.

What makes this finding particularly compelling is the combination of statistics and the specificity of the candidate particle. Physicists have had glueball suspects before, but previous evidence was always murky enough to explain away. The new BES III results push past that ambiguity in ways that are difficult to dismiss — though the collaboration is being careful not to claim a definitive discovery just yet.

The Standard Model of particle physics already survived one major vindication with the Higgs boson confirmation in 2012. Glueballs represent a different kind of test — not a missing piece of the framework, but a confirmation that the framework's description of the strong force is genuinely correct in a regime that's notoriously hard to probe experimentally.

If the glueball result holds up under further scrutiny, it would be one of the more satisfying moments in recent physics history. Not a revolution, but a deep confirmation that the universe really does work the way the math says it should — right down to particles made of nothing but the force that holds everything else together.