Join our Newsletter — 33% off our NHI Course
Interactive report · updated live

250+ NHI & AI Agent Breaches

We have analysed 250 real-world breaches involving non-human identities and AI agents — leaked API keys, stolen tokens, compromised service accounts and the new AI attack paths. Explore how attackers get in, what they take, and what to fix first.

Explore every breach
250Breaches analysed2017–2026
46%Credential-ledLeaked, stolen or weak credentials
Secret or passwordMost-exploited identityIn 98 breaches
76AI-relatedAI agents, LLM platforms, prompt injection
What the data tells us

The Patterns Keep Repeating

Calculated live from the breaches below. Hover or tap a card for what to do about it.

23%

Supply chain is the #1 known way in — 58 of 250 breaches.

Hover or tap — what to do

What to Do

Attackers reuse what works. Prioritise controls against the most common entry route first, then work down the chart.

See these breaches
46%

114 breaches began with a leaked, stolen or weak credential.

Hover or tap — what to do

What to Do

Scan code, CI logs and tickets for secrets, vault what you find, rotate on exposure and replace long-lived keys with short-lived, scoped credentials.

See leaked-secret breaches
30%

74 breaches came through the supply chain or a SaaS / OAuth integration.

Hover or tap — what to do

What to Do

Your NHI risk includes your vendors' NHIs. Inventory third-party OAuth grants and integrations, scope them down and revoke what is unused.

See supply-chain breaches
76

breaches involve AI agents, LLM platforms or AI-specific attack routes.

Hover or tap — what to do

What to Do

Treat every agent as a governed identity: an owner, least-privilege and short-lived credentials, tool allow-lists, approval for risky actions and a kill switch.

See AI-related breaches
The breakdown

How Attackers Get In

A breach can have more than one route or identity type, so bars can add up to more than 250.

By Year

By Attack Route

By Identity Type

By Category

Click any bar to filter the breach list. 23 breaches (9%) did not disclose how the attacker got in.

All breaches

Explore Every Breach

Showing 24 of 250 matching breachesTimeline view

Nov 2026 NHI

Wyndham Breach 2026: How a Password Hardcoded in an Android App Exposed Customer Login Tokens

A password hardcoded in Wyndham's Android app opened a Salesforce-hosted database of customer data and login tokens, TechCrunch reports. No misuse confirmed.

Leaked secret · Secret or password · Session token
Nov 2026 NHIAI agents

ARTEX Agentic Secret Harvesting 2026: How AI Agents Turned Leaked GitHub Secrets into Key Vault Theft

Microsoft says ARTEX AI agents found leaked GitHub secrets, signed in as service principals across tenants and exfiltrated Key Vault secrets in five minutes.

Leaked secret · OAuth / SaaS integration · Service account · Secret or password · OAuth token · Cloud credential · API key
Nov 2026 NHIAI agents

Claude Code Settings Leak 2026: How “Don’t Ask Again” Approvals Put 2,539 Secrets in Public Repos

Apiiro found Claude Code's "don't ask again" approvals saved database passwords, tokens and AWS keys into files in 1,994 public repos. No abuse confirmed.

Leaked secret · Misconfiguration · Secret or password · Source control token · Cloud credential · AI agent
Nov 2026 NHI

Chainstack Breach 2026: How a Third-Party Component Led to Exposed Node Credentials and API Keys

An attacker entered Chainstack via a third-party component, moved from dev to production and reached systems holding node credentials. Rotate keys.

Supply chain · API key · Secret or password
Nov 2026 NHIAI agents

GitBook Breach 2026: How an AI-Assisted Attack Reached API Tokens and Git Sync Secrets

GitBook breach 2026: an SSO flaw and a misconfigured GitHub App led to a Cloudflare key, production secrets and customer API tokens. What happened and why.

Vulnerability exploit · Stolen credentials · Source control token · API key · Service account · Secret or password
Nov 2026 AI agentsNHI

WorkOS Breach 2026: How an Escaped AI Agent Used a Development Credential Left in Production

A researcher's ARTEX AI agent escaped its sandbox and used a predictable dev credential left in production to bulk-pull WorkOS sales and support data.

Weak or default credentials · AI agent misbehaviour · Secret or password · AI agent
Oct 2026 NHILLM & AI platform

Northflank Breach 2026: How Customers’ AI API Keys Were Abused After Environment Variables Were Exposed

Northflank told customers to assume all environment variables were decrypted after a Wiz alert, then confirmed abuse of AI API keys stored on its platform.

Not disclosed · API key · Secret or password
Oct 2026 NHI

Applebee’s Breach 2026: How a Database Password Left in the Android App Exposed Customer Data

A database password hardcoded in Applebee's Android app exposed customer names, birthdates, phones and IPs, TechCrunch reports. Fixed 29 Oct; no misuse found.

Leaked secret · Secret or password
Oct 2026 AI agentsNHI

Chainalysis Breach 2026: How an Agentic Intrusion Exposed Live Tokens and Password Hashes for All Users

An attacker using offensive agentic techniques exposed Chainalysis users' password hashes and live tokens, forcing API key rotation and KYT downtime.

Not disclosed · Session token · API key
Oct 2026 NHI

Binance Hack 2026: How a Compromised DevOps Laptop Led to $6.3 Billion in Unauthorised Withdrawals

Binance says a compromised DevOps laptop let an attacker reach a service container and submit $6.3 billion in unauthorised withdrawals. What we know so far.

Not disclosed · Service account
Oct 2026 NHIAI agents

Cursor Marketplace Extension Compromise 2026: How a Tampered Claude Code Update Planted an Attacker SSH Key

A tampered Claude Code extension served via Cursor's marketplace added an attacker SSH key and reverse tunnel for up to 5,000 users. What to do next.

Supply chain · CI/CD or publishing token
Oct 2026 NHIAI agents

Vapor Panda ARTEX Campaign 2026: How AI Agents Forged Admin Tokens Across 200 Organisations

CrowdStrike says China-linked Vapor Panda used ARTEX AI agents against 200+ organisations, forging an admin token from old mail-server keys. Lessons.

Vulnerability exploit · Stolen credentials · Signing key or certificate · Session token · Cloud credential
Oct 2026 NHILLM & AI platform

LiteLLM Supply Chain Attack 2026: How Stolen GitHub OAuth Tokens Shipped a Credential Stealer to PyPI and Docker

LiteLLM 1.107.0 was poisoned on 21 October 2026 using three engineers' stolen GitHub CLI OAuth tokens. What was stolen, who is affected and what to rotate.

Supply chain · Stolen credentials · OAuth token · CI/CD or publishing token · Cloud credential · API key
Oct 2026 NHILLM & AI platform

Railway Breach 2026: How an Attacker Decrypted Customer Environment Variables to Hunt AI Keys

In October 2026 an attacker used stolen Railway CI credentials to decrypt customer environment variables in search of AI provider keys. Cause undisclosed.

Not disclosed · Secret or password · API key
Oct 2026 NHI

OVHcloud CLI Compromise 2026: How a Hijacked GitHub Account Swapped Release Binaries for a Secret Stealer

A hijacked contributor GitHub account replaced OVHcloud CLI v0.11.0 to v0.15.0 release binaries with a stealer that hunts secrets on disk. What it means.

Supply chain · Not disclosed
Oct 2026 NHI

UpdraftCentral Key Theft 2026: How Stolen Connection Keys Were Used to Backdoor WordPress Sites

Keys stolen from TeamUpdraft's customer database signed UpdraftCentral commands that installed 99 backdoor plugins on WordPress sites. What to do.

Stolen credentials · Supply chain · Signing key or certificate · API key
Oct 2026 NHI

Zeabur Second Breach 2026: How Unrotated Admin Tokens Exposed 3.2 Million Environment Variables

How non-expiring Zeabur admin tokens missed after the August breach let an attacker read 3.2M environment variables and 321,810 accounts in October 2026.

Stolen credentials · Leaked secret · Secret or password · Cloud credential · Service account · API key
Oct 2026 NHI

TiDB Cloud Breach 2026: How Stolen Service Account Credentials Exposed PingCAP Customer Data and Source Code

PingCAP says stolen service account credentials let an attacker copy TiDB Cloud customer records, source code and encrypted API keys over four weeks in 2026.

Stolen credentials · Service account · API key · Cloud credential
Oct 2026 NHILLM & AI platform

OpenCode Console Breach 2026: How Stored OpenAI, Anthropic and Google API Keys Were Stolen and Abused

A legacy OpenCode system exposed stored OpenAI, Anthropic and Google keys for 1.4% of Console users in October 2026, leading to unauthorised AI charges.

Not disclosed · API key
Oct 2026 NHI

ConnectWise Cloud Backup Breach 2026: How a Stolen App Secret Opened 160+ Microsoft 365 Tenants

A stolen ConnectWise Cloud Backup app credential let attackers sign in to 160+ Microsoft 365 tenants with no MFA, then self-assign admin roles.

OAuth / SaaS integration · Stolen credentials · Service account · Secret or password
Oct 2026 NHILLM & AI platform

Fireworks AI Breach 2026: How Stolen Internal Credentials Exposed Customer API Keys in Evaluation Jobs

Fireworks AI internal credentials and a third-party cloud compromise exposed customer API keys, secrets and datasets, with Notion AI chats hit downstream.

Stolen credentials · Supply chain · API key · Secret or password · Service account
Oct 2026 NHI

Tensorlake npm Worm 2026: How a Shai-Hulud Stealer Turned Registry Tokens Into Signed Malicious Releases

tensorlake 0.5.144 shipped a Shai-Hulud style worm via the project's own pipeline, stealing npm, GitHub and cloud tokens and republishing as its victims.

Supply chain · CI/CD or publishing token · Source control token · Cloud credential · Secret or password
Oct 2026 NHI

GhostAction Returns 2026: How Stolen GitHub Credentials Planted Secret-Stealing Workflows in 772 Repositories

GitGuardian found GhostAction's fake security workflow pushed to 772 repos in Aug-Sept 2026 using stolen GitHub credentials, targeting 2,577 CI/CD secrets.

Stolen credentials · Supply chain · Source control token · Cloud credential · CI/CD or publishing token · Secret or password
Sep 2026 AI agentsNHI

OpenAI Agents and US Government Websites 2026: How Rogue AI Agents Used Leaked Census API Keys and Probed Federal Sites

OpenAI agents used Census API keys leaked on GitHub, reposted SEC data and, per Transluce, tried to hack an Education site. No compromise found.

AI agent misbehaviour · Leaked secret · AI agent · API key

Don't Be the Next Entry

Find your exposed NHIs and AI agents before attackers do — with a risk assessment or our training.

Take the report with you. The patterns, how attackers get in, the AI agent breaches, twelve breaches and their lessons, and a 30 and 90 day plan, in a nine-page PDF.

PDF Downloads for Newsletter Subscribers

Get the guide

Enter your details and we’ll email you a download link.

The full guide is free to read online. The PDF is a thank you for subscribing.

Fill in all fields marked * and tick the PDF box to continue.