2,500 companies just had their AI infrastructure exposed. The FBI is involved. Your CI/CD pipeline is likely next.

TeamPCP just pulled off the largest AI supply chain breach of 2026.

40 minutes.

That's all it took.

A compromised version of LiteLLM sat on PyPI for 40 minutes in March.

In that window, 434,000 CI/CD pipelines ran the poisoned package.

2,500+ organizations had their credentials harvested.

Cisco alone: 327 secrets, 1,900 pipeline runs exposed.

X Corp (Twitter): 3,459 secrets.

Deloitte: 462 secrets, 503 runs.

AWS, Samsung, Salesforce, Siemens, FedEx, Volkswagen — all high-confidence exposures.

The FBI issued FLASH advisory FLASH-20260702-01 in July warning that attackers are likely weaponizing these credentials right now.

Here's what makes this different from every other breach.

This wasn't a hack of your infrastructure.

It was a hack of your AI tool's build pipeline.

LiteLLM never got directly attacked. Its CI server installed Trivy, a vulnerability scanner, from the system package manager. Trivy had been compromised weeks earlier through a leaked automation token that was rotated but never fully revoked.

One un-revoked token. Three tools deep. That's all it took.

The malware ran on every Python invocation. No import required. It scraped SSH keys, cloud credentials, Kubernetes tokens, LLM API keys — everything the process could touch.

Then it shipped the loot to a typosquatted domain. Where exfiltration failed, it uploaded stolen data to the victim's own GitHub as a public release asset.

Your AI layer is now the most dangerous attack surface in your stack.

Rotate every credential your CI/CD pipeline ever touched. Pin your dependencies. Assume your AI proxy keys are compromised. Audit your LiteLLM versions 1.82.7 and 1.82.8 immediately.

SOURCE: https://www.securityweek.com/over-2500-organizations-impacted-by-litellm-supply-chain-attack/
VERIFIED: SecurityWeek, CloudSEK, Ars Technica
SIGNAL: This is the first major supply chain breach targeting AI infrastructure specifically. The FBI is treating it as an active threat. Your CISO needs this briefing today.