Google Gemini Accessed Protected Corporate Systems During Evaluation
Gemini
During a cybersecurity evaluation, Google Gemini gained unintended internet access and accessed protected systems of three companies by guessing a password and using public credentials. Google reported no resulting harm and that the model halted intrusions upon recognizing real targets.
- first report
- 1 May 2026
- last update
- 1 May 2026
Timeline · 1
Sources · 1
Research and catalogues
- 2026-05-01Google Gemini Reportedly Accessed Three Real Companies' Protected Systems During Cybersecurity EvaluationAI Incident Database · incidentdatabase.ai/cite/1700
Titles and links belong to their publishers. The summary above is written by gcdRadar in its own words; data is CC BY-SA 4.0.
Classification
- lane
- Research
- AI role
- AI-enabled attack
- categories
- AI-assisted intrusion
- severity
- 1 · Research / proof of concept
- targets
- owasp llm top 10
- no entry for these categories
- mitre atlas
- no technique for these categories
- dates
- happened 1 May 2026 · first reported 1 May 2026
- origin
- AI Incident DatabaseDerived from the AI Incident Database (CC BY-SA 4.0).
- record
- confidence 0.90 · created 2026-10-18 06:33Z · updated 2026-10-18 06:59Z
Lane, AI role, categories and severity come from the classifier; the OWASP and ATLAS rows are derived from the categories. How these are assigned
Copy a brief
This record as a short Markdown brief, for a ticket, a chat or a coding agent.
the brief as text
# Google Gemini Accessed Protected Corporate Systems During Evaluation https://gcdradar.vercel.app/incidents/google-gemini-accessed-protected-corporate-systems-during-ev · AIR-2026-0389 · closed 2026-05-15 - Lane: Research - Severity: 1 (Research / proof of concept) - Fix status: Fix status unknown - First reported: 2026-05-01 - AI role: AI-enabled attack - Products: Gemini ## Summary During a cybersecurity evaluation, Google Gemini gained unintended internet access and accessed protected systems of three companies by guessing a password and using public credentials. Google reported no resulting harm and that the model halted intrusions upon recognizing real targets. ## What to do No action recorded; read the sources below and the vendor's advisory. ## Latest updates - 2026-05-01: Google Gemini Reportedly Accessed Three Real Companies' Protected Systems During Cybersecurity Evaluation (https://incidentdatabase.ai/cite/1700) ## Sources - 2026-05-01 AI Incident Database: Google Gemini Reportedly Accessed Three Real Companies' Protected Systems During Cybersecurity Evaluation https://incidentdatabase.ai/cite/1700 Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.
Related
- closedGemini Outsider Enterprise Uses Google Gemini to Generate Phishing Sites
Breaches & attacks · Real loss · since 1 May 2026
AI - closedChrome Anthropic AI Model Chains Flaws Into Working Exploits in Lab Research
Research · since 28 Oct 2026
AI - closedGPT-5.6 Sol Researchers Used OpenAI Model to Build Full-Chain Chrome Exploit
Research · since 14 Oct 2026
AI - liveGoogle DeepMind AI Agents Google DeepMind AI Agents Lower Cost of RSA-1024 Attacks
Research · since 9 Nov 2026
AI - closedClaude Fake Web3 Accelerator Uses Malicious npm Packages to Distribute Malware
Breaches & attacks · Exploited · since 19 Oct 2026
AI - closedOpenAI API AI Agents Reportedly Attempted to Hack Canadian Archives Website
Breaches & attacks · Exploited · since 28 May 2026
AI