AI vs AI: Agent hacked McKinsey's chatbot and gained full read-write access in just two hours
2026-03-10T07:24:03Z•4c04f0171006e98643388c751ac0de997bfa81f0b4824f4fc64f091338d65ebd
AI-agent compromiseCisco SD‑WANReact2Shellactive exploitationchatbot securitycredential/API key theftdata breachinformation-stealernation-state activityphishingprivacyransomwarespywaresupply-chain/open-source abuse
What happened
A broad cluster of high-impact security incidents and threat trends reported across multiple organisations and sectors. Key items: an AI agent demonstration that compromised McKinsey’s internal chatbot with full read/write access; multiple large data breaches and ransomware incidents (ShinyHunters’ claimed Salesforce customer data heist, LexisNexis Legal & Professional breach, ELECQ ransomware exposing customer contact data); nation-state and state‑linked activity (Iran-linked backdoors in US networks, Russian-linked phishing targeting Signal/WhatsApp, Hamas-linked spyware disguised as an EU/
Why it matters
A reviewed impact interpretation has not been published for this record.
Evidence and limitations
- Source ID
- theregister_security
- Record identifier
- 4c04f0171006e98643388c751ac0de997bfa81f0b4824f4fc64f091338d65ebd
- Enrichment time
- 2026-03-10T07:24:03Z
- AI-assisted enrichment
- Yes
This record may overlap with other records. Its enrichment can be incomplete or wrong, and machine assistance was used. Validate consequential decisions against the linked source and your own environment.