
Railway says stolen customer AI keys drove $740,000 in unauthorized charges
An attacker decrypted secrets in at most about 100,000 workspaces; Railway will cover some charges providers will not refund.
37 stories, newest first · one card per story

An attacker decrypted secrets in at most about 100,000 workspaces; Railway will cover some charges providers will not refund.

Apiiro research exposed 2,539 secrets across 1,994 public GitHub repositories via Claude Code configuration files.
Apache Airflow audit log stored Connection and Variable secret values in cleartext because the secrets masker matched only on configured sensitive key names.
Different secret ids can collide in multi-team deployments, disclosing variables or connections.
Apiiro found 2,539 real secrets in 1,994 public GitHub repositories.

Coordinated attack on 61 repositories used command injection in GitHub Actions to steal CI/CD secrets; 83 malicious issues and PRs identified.

The company says it blocked the exploit and emailed affected account owners mitigation steps.

Workflows on more than 13,500 branches logged secrets; public runs were held, while private and Enterprise Server runs completed.

Wiz found over 13,500 planted branches from more than 850 accounts; GitHub held public runs but not private ones.

The post contrasts first-party fixes with third-party upgrades and warns agents often pull outdated dependencies.

October 10 and 12 workflows posted CI secrets to webhook.site; a key committed May 27 was never revoked.

An unauthorized party may have decrypted some environment variables; deployments stay paused while builds return in stages.

Hackers may have decrypted environment variables for a subset of users; Railway paused deployments while it investigates.

A subset of Railway users may have had decrypted API keys and credentials exposed; the company urges rotation and has revoked its own tokens.

Deployments are paused, and Railway told every user to rotate API keys and registry credentials while it identifies affected accounts.

Researcher says ChatGPT’s GitHub connector can be prompt-injected into stealing private repository secrets.

Trellix says normal AI-agent workflows can leak API keys and cloud secrets into logs, transcripts, and model context.

Trellix warns AI agent workflows can leak credentials from vaults into commands, logs, and model context.

GhostAction compromised about 1,900 GitHub accounts and pushed malicious workflows to steal CI secrets from tens of thousands of repositories.

GhostAction compromised maintainer accounts to plant credential-stealing GitHub Actions workflows in tens of thousands of repositories.

GhostAction hit 346 GitHub repositories, including Uber's, with malicious Actions workflows that stole CI secrets.

Malicious tensorlake@0.5.144 on npm delivers a Shai-Hulud worm that steals developer secrets and can wipe hosts.

GhostAction hit 772 GitHub repositories with malicious Actions workflows that stole CI/CD secrets.
Common secret-wiping patterns can add stack copies or erase memory that never held the secret.
Today 1.5K readers · 2.4K views · 753 reads · all time 189K views