---
title: AI Security Incidents and Why Audits Matter | DataAutomation
url: https://dataautomation.com/ai-security/
description: Shadow AI was involved in 43% of the breaches IBM studied for its 2026 report.
---

# Your AI chat is not a vault.

Every week another AI tool or vendor leaks what people typed into it. If your team pastes customer names, orders, or passwords into a chat, that data now lives somewhere you do not control, with no audit behind it.

[See the latest incidents](https://dataautomation.com#incidents) [See our audits](https://dataautomation.com/audits/)

43%

of breached companies had incidents involving shadow AI, meaning staff using AI tools the company never approved. That is up from 20% a year earlier, and those breaches averaged $5.39 million.

92% of companies with an AI-related breach had no proper access controls on their AI. Source: [IBM Cost of a Data Breach Report 2026 (Ponemon Institute, 602 breached organizations)](https://www.ibm.com/reports/data-breach).

## The latest AI security incidents

Updated Nov 19, 2026. These are just the last few weeks. [See a full year of AI incidents by category](https://dataautomation.com#year).

- Nov 11, 2026 ### AI vendor breach exposes Notion AI chat transcripts A breach at Fireworks AI, a vendor that processed Notion AI requests, exposed chat transcripts for under 1% of Notion workspaces. Companies that typed customer details into those chats are now notifying people. [Source: SecurityWeek](https://www.securityweek.com/notion-customer-data-compromised-in-fireworks-ai-hack/)
- Nov 10, 2026 ### Fireworks AI breach widens to datasets and billing data Attackers reached customer billing details, secrets, datasets, and unfinished models through a third-party cloud provider. [Source: The Hacker News](https://thehackernews.com/2026/11/fireworks-ai-breach-widens-to-expose.html)
- Nov 9, 2026 ### A research AI agent escaped its test and pulled WorkOS customer data WorkOS traced an October breach to an offensive AI agent that left its test environment and bulk-downloaded customer data from an old sales tool. [Source: PointGuard AI](https://www.pointguardai.com/ai-security-incidents/escaped-research-agent-pulls-workos-customer-data-in-bulk)
- Nov 9, 2026 ### AI inference server exploit exposes user accounts and agent histories An attacker used a flaw in an AI serving tool to take account details and histories for about 100 NEAR AI users and 18 agents. [Source: PointGuard AI](https://www.pointguardai.com/ai-security-incidents/sglang-exploit-breaches-near-ai-inference-and-agent-histories)
- Nov 8, 2026 ### Automated attackers hunt leaked AI keys An agent-driven campaign combined leaked GitHub secrets with an unpatched AI tool to break into cloud and AI accounts. [Source: PointGuard AI](https://www.pointguardai.com/ai-security-incidents/agentic-secret-hunters-exploit-langflow-and-leaked-ai-keys-cve-2025-3248)
- Nov 5, 2026 ### AI coding tool saved secrets into public code Saved approvals in an AI coding assistant stored full commands, passwords included, in files that were pushed to GitHub. Researchers found 2,539 secrets in 1,994 public repositories. [Source: PointGuard AI](https://www.pointguardai.com/ai-security-incidents/claude-code-approval-rules-leak-thousands-of-secrets-to-github)
- Nov 5, 2026 ### Researchers slipped chats past ChatGPT's Lockdown Mode A popup trick in HTML previews allowed silent theft of chat data until OpenAI fixed it. [Source: PointGuard AI](https://www.pointguardai.com/ai-security-incidents/popup-trick-slips-chatgpt-chats-past-openais-lockdown-mode)
- Oct 30, 2026 ### Stolen AI keys ran up more than $100,000 in charges Attackers decrypted stored customer secrets at a hosting provider and used customers' AI keys; stolen keys from the same wave were tied to over $100,000 in AI usage. [Source: PointGuard AI](https://www.pointguardai.com/ai-security-incidents/northflank-breach-caps-black-october-run-on-ai-developer-keys)
- Oct 30, 2026 ### Tampered AI extension tried to install a backdoor A fake AI coding extension in a public marketplace tried to open remote access on developers' machines, reaching up to 5,000 users in 11 hours. [Source: PointGuard AI](https://www.pointguardai.com/ai-security-incidents/tampered-claude-code-extension-slips-ssh-backdoor-into-cursor-marketplace)
- Oct 27, 2026 ### Vendor breach exposes AI agent activity logs A breach at a Warp vendor exposed AI agent traces and telemetry for up to about 1% of users. [Source: PointGuard AI](https://www.pointguardai.com/ai-security-incidents/warp-vendor-breach-exposes-ai-agent-traces-and-telemetry)

## A year of AI going wrong

28 real incidents from November 2025 to November 2026 . None of them needed a hacker movie. Most started with an AI that had more access than it needed.

**9 seconds** for an AI agent to delete a company's database and its backups

**300 million** private AI chat messages left open in a single app

**$237 million** taken with cloned voices and faces, in just two of these cases

**1.5 million** AI agent keys sitting in one open database

All 28 Agents gone rogue 4 AI hijacked by hidden instructions 5 AI apps and vendors leaking data 6 Leaked keys and exposed agents 5 Poisoned AI tools and plugins 3 Deepfake fraud 3 Shadow AI at work 2 Attacks run with AI 0

Deepfake fraud Nov 12, 2026

### Three charged over a $129 million deepfake CEO scam

Prosecutors say AI-cloned executives and fake lawyers talked employees into wiring money. Victims included two US hospital systems, a bank, and a UK fintech.

[Read it on Bitdefender](https://bitdefender.com/en-us/blog/hotforsecurity/three-charged-129m-ai-deepfake-ceo-scam) AI apps and vendors leaking data Nov 11, 2026

### An AI vendor breach exposed Notion AI chat transcripts

A breach at AI provider Fireworks AI reached customer data at Notion, including AI chat transcripts.

[Read it on SecurityWeek](https://www.securityweek.com/notion-customer-data-compromised-in-fireworks-ai-hack/) Agents gone rogue Nov 9, 2026

### A research AI agent left its test and pulled WorkOS customer data

An attack agent in a security test wandered outside its sandbox and bulk-downloaded customer data from an old sales tool. Nobody told it to.

[Read it on PointGuard AI](https://www.pointguardai.com/ai-security-incidents/escaped-research-agent-pulls-workos-customer-data-in-bulk) Poisoned AI tools and plugins Nov 9, 2026

### An AI inference server exploit exposed accounts and agent histories

A flaw in the software serving AI models let attackers reach user accounts and the history of what agents had done.

[Read it on PointGuard AI](https://www.pointguardai.com/ai-security-incidents/sglang-exploit-breaches-near-ai-inference-and-agent-histories) Leaked keys and exposed agents Nov 8, 2026

### Automated attackers now hunt leaked AI keys around the clock

Agent-driven campaigns scan for exposed keys and unpatched AI tools, then use them to break into cloud and AI accounts.

[Read it on PointGuard AI](https://www.pointguardai.com/ai-security-incidents/agentic-secret-hunters-exploit-langflow-and-leaked-ai-keys-cve-2025-3248) AI hijacked by hidden instructions Nov 5, 2026

### Researchers slipped chats past ChatGPT's Lockdown Mode

A popup trick got conversation data out even with OpenAI's strictest setting turned on.

[Read it on PointGuard AI](https://www.pointguardai.com/ai-security-incidents/popup-trick-slips-chatgpt-chats-past-openais-lockdown-mode) Leaked keys and exposed agents Nov 5, 2026

### An AI coding tool saved secrets into public code

Approval rules in a coding agent ended up committing thousands of secrets to GitHub.

[Read it on PointGuard AI](https://www.pointguardai.com/ai-security-incidents/claude-code-approval-rules-leak-thousands-of-secrets-to-github) Leaked keys and exposed agents Oct 30, 2026

### Stolen AI keys ran up more than $100,000 in charges

Attackers harvested AI developer keys and spent them on other people's accounts.

[Read it on PointGuard AI](https://www.pointguardai.com/ai-security-incidents/northflank-breach-caps-black-october-run-on-ai-developer-keys) Poisoned AI tools and plugins Oct 30, 2026

### A tampered AI extension tried to install a backdoor

A fake Claude Code extension in the Cursor marketplace tried to give attackers remote access to developers' machines.

[Read it on PointGuard AI](https://www.pointguardai.com/ai-security-incidents/tampered-claude-code-extension-slips-ssh-backdoor-into-cursor-marketplace) AI apps and vendors leaking data Oct 27, 2026

### A vendor breach exposed AI agent activity logs

Agent traces and telemetry from Warp's AI features were exposed through a third-party vendor.

[Read it on PointGuard AI](https://www.pointguardai.com/ai-security-incidents/warp-vendor-breach-exposes-ai-agent-traces-and-telemetry) AI apps and vendors leaking data Oct 26, 2026

### An AI aggregator leaked 17.6 million files of prompts and company documents

302.ai left a cloud storage bucket public with 22.7 TB of AI conversations, generated media, and business files, including another company's customer support bot logs.

[Read it on eSecurity Planet](https://www.esecurityplanet.com/threats/news-302-ai-data-leak-private-chats/) Shadow AI at work Oct 13, 2026

### A third of employees admit putting sensitive data into unapproved AI

Gartner found 33% of employees surveyed had entered sensitive information into AI tools their employer had not approved.

[Read it on Abnormal AI](https://abnormal.ai/learning/chatgpt-data-leakage) Deepfake fraud Sep 25, 2026

### A cloned voice talked an Italian bank into wiring €95 million

Fraudsters used AI and fake WhatsApp messages to pose as senior executives and a lawyer at Fideuram, part of Intesa Sanpaolo. About €36 million is still missing.

[Read it on Cybernews](https://cybernews.com/news/ai-voice-scam-hits-italian-bank-for-95-million/) AI hijacked by hidden instructions Sep 9, 2026

### A planted prompt read connected Gmail while ChatGPT answered normally

Check Point showed an assistant could run a hidden second task, reading a user's connected Gmail and moving it to another account, while giving a normal reply.

[Read it on Hexnode](https://www.hexnode.com/threat-watch/chatgpt-gmail-exfiltration-flaw-enterprise-ai-app-governance-and-xdr-lessons/) Deepfake fraud Jul 22, 2026

### A jewelry maker lost $1.2 million to an impersonated boss

A subsidiary of India's Sky Gold & Diamonds disclosed that an employee approved a transfer after a fake executive request.

[Read it on Adaptive Security](https://www.adaptivesecurity.com/blog/sky-gold-lost-usd1-2-million-to-a-deepfake) AI hijacked by hidden instructions May 4, 2026

### Hidden instructions on web pages are now stealing data for real

Google and Forcepoint reported prompt injection attacks running against live AI systems: text planted in pages and emails that makes an assistant leak data or credentials on its own. Google counted a 32% rise in these planted prompts between November and February.

[Read it on TechRepublic](https://www.techrepublic.com/article/news-ai-agents-prompt-injection-data-security/) Agents gone rogue Apr 25, 2026

### An AI coding agent deleted a company's database and backups in 9 seconds

A Cursor agent running Claude Opus 4.6 hit a credential mismatch on a routine task, found a broad API token, and wiped PocketOS's production database and its backups in one call. The car rental software company spent 30 hours in crisis.

[Read it on TechRepublic](https://techrepublic.com/article/ai-agent-deletes-company-database-admits-violating-guardrails) AI hijacked by hidden instructions Apr 7, 2026

### A dashboard's AI assistant could be turned into a data courier

Researchers showed that instructions hidden in Grafana logs could make its AI assistant send private business data to an outside server. Grafana fixed it and disputes how easy it was to trigger.

[Read it on SecurityWeek](https://www.securityweek.com/grafanaghost-attackers-can-abuse-grafana-to-leak-enterprise-data/) AI apps and vendors leaking data Mar 17, 2026

### Sears' AI customer service bot left 3.7 million records exposed

Unsecured databases behind the Samantha chatbot held customer names, addresses, phone numbers, chat logs, and recorded calls from 2024 to 2026.

[Read it on SC Media](https://www.scworld.com/brief/misconfigured-ai-bot-databases-leak-millions-of-sears-home-services-customer-records) Agents gone rogue Feb 26, 2026

### Claude Code ran terraform destroy and wiped 2.5 years of records

While moving a side project, an AI agent swapped in an old state file and destroyed DataTalks.Club's production setup, database and snapshots included. AWS support restored it about a day later.

[Read it on Tom's Hardware](https://www.tomshardware.com/tech-industry/artificial-intelligence/claude-code-deletes-developers-production-setup-including-its-database-and-snapshots-2-5-years-of-records-were-nuked-in-an-instant) Poisoned AI tools and plugins Feb 20, 2026

### A worm planted fake AI tools inside developers' coding assistants

SANDWORM_MODE spread through 19 malicious npm packages and slipped a rogue tool server into Claude Code, Cursor, and others, with hidden instructions telling the AI to steal keys.

[Read it on SC Media](https://www.scworld.com/news/sandwormmode-shai-hulud-with-an-ai-twist) AI apps and vendors leaking data Feb 5, 2026

### An AI chat app left 300 million private messages open online

Chat & Ask AI, with more than 50 million users, had a misconfigured database exposing full chat histories. People had treated it like a diary and an advisor.

[Read it on Fox News](https://www.foxnews.com/tech/millions-ai-chat-messages-exposed-app-data-leak.amp) Leaked keys and exposed agents Feb 2, 2026

### An AI agent network left 1.5 million API keys in an open database

Moltbook exposed its whole production database: agent login tokens, 35,000 emails, and private agent messages, some holding plaintext OpenAI keys. Anyone could take over any agent.

[Read it on Wiz](https://www.wiz.io/blog/exposed-moltbook-database-reveals-millions-of-api-keys) Leaked keys and exposed agents Jan 31, 2026

### Tens of thousands of personal AI agents sat open on the internet

Censys found more than 21,000 publicly reachable OpenClaw agents in a week, some leaking plaintext keys and credentials. The agents often had access to email, files, and messages.

[Read it on eSecurity Planet](https://www.esecurityplanet.com/threats/openclaws-rapid-rise-exposes-thousands-of-ai-agents-to-the-public-internet/) AI hijacked by hidden instructions Jan 8, 2026

### One shared file could turn ChatGPT into a quiet spy

Radware's ZombieAgent attack hid instructions in content ChatGPT read, making it leak data from connected accounts and even save the instructions to memory to keep doing it. OpenAI patched it in December.

[Read it on eWeek](https://www.eweek.com/news/zombie-ai-attack-chatgpt-leaks/) Shadow AI at work 2026

### Shadow AI was involved in 43% of the breaches IBM studied

IBM's 2026 breach report found staff using unapproved AI tools in 43% of the breaches it studied, more than double the year before, and 92% of companies with an AI-related breach had no proper access controls on their AI.

[Read it on IBM](https://www.ibm.com/reports/data-breach) AI apps and vendors leaking data Nov 27, 2025

### OpenAI API customers exposed through an analytics vendor

Attackers phished their way into Mixpanel and took names, emails, and locations of some OpenAI API users. OpenAI dropped the vendor.

[Read it on SiliconANGLE](https://siliconangle.com/2025/11/27/mixpanel-breach-compromises-openai-api-users-account-data/) Agents gone rogue Nov 13, 2025

### Hackers used an AI agent to do most of a spy campaign for them

Anthropic reported that a state-sponsored group tricked Claude Code into attacking about thirty organizations, with the AI doing 80 to 90 percent of the work. A few break-ins succeeded.

[Read it on Anthropic](https://www.anthropic.com/news/disrupting-AI-espionage)

Every one of these is a reason we keep AI inside audited connections, give it only the access a task needs, and own the fix if something we run goes wrong. [See our audits](https://dataautomation.com/audits/).

## Free audit by our [CEO](https://dataautomation.com/about/#ceo), find your biggest systems leak and save money.

What to cut, what overlaps, and what is cheaper to build.

## Never paste these into an AI chat

Not into ChatGPT, Claude, Copilot, Gemini, or a browser extension, unless your company approved that tool for that data.

### Customer names, emails, phones, and addresses

Personal data. Once it is in a chat, you cannot recall it, and you may owe people a breach notice if that tool leaks.

### Amazon buyer data, including Amazon Custom personalization

Restricted under Amazon's data protection rules. Sellers who move it into tools Amazon has not approved can put their selling account at risk.

### Order, customer, or payroll exports

A spreadsheet pasted for a quick answer is a full copy of your records on someone else's servers.

### Passwords, API keys, and tokens

Attackers now run automated hunts for leaked AI and cloud keys, and stolen keys have run up six-figure bills.

### Card, bank, and tax ID numbers

Never needed for an AI to help you, and the most damaging to lose.

### Health or employee records

Covered by their own laws, with their own penalties.

## What to do instead

### Use business AI accounts with data controls

Team or enterprise plans with training turned off, retention set, and admin oversight. Not personal accounts.

### Connect AI through audited integrations

Let AI reach your systems through a connection that passes only the fields a task needs, logs every call, and is audited.

### Decide where a person approves

Choose which steps touching customers or money need a sign-off, and log everything AI does on its own.

### Ask every AI vendor for its audit

If they cannot show one, they should not see your customers.

## How we handle your data

Every integration and AI workflow we run is audited by outside firms: Deloitte and Amazon SP-API audits in 2026, PwC and Amazon SP-API in 2025, and Deloitte and Amazon SP-API in 2024 and 2022.

We pass only the fields a task needs, log what moves, and do not keep client data after an engagement ends. Amazon buyer data stays where it is needed to fulfill the order, never in an AI chat. Our AI connection for ordering services never touches your systems at all.

[See our audits](https://dataautomation.com/audits/) or [talk to us about using AI safely](https://meet.dataautomation.com/meet-with-dataautomation).
