AI safety failures
Unsafe deployments, ignored test results, model behavior that could harm the public, or internal warnings that were minimized.
Public investigative tip line
Know SomethingAbout AI Harm?
ChatGPTAiML Tips is collecting serious leads, documents, screenshots, and firsthand information about AI risks, misuse, hidden systems, breaches, privacy violations, data center expansion, and public harm.
Submit if you have evidence of
AI systems affecting the public without disclosure
Internal warnings about safety, privacy, or security
Documents, screenshots, logs, memos, or contracts
Environmental, labor, or community impacts being hidden
Corporate or government AI decisions people cannot challenge
Do not ignore the detail that feels too specific.Names, dates, screenshots, contracts, tickets, permits, logs.
What we are looking for
Tips That Can Move a Story Forward
We are looking for information that helps verify what is happening, who is responsible, and how the public, workers, customers, communities, or institutions are affected.
Data center impact
Power contracts, water use, land pressure, environmental permits, community opposition, grid strain, or hidden expansion plans.
Security breaches
Exposed prompts, leaked API keys, compromised model systems, vendor incidents, phishing, or misuse of AI in attacks.
Privacy violations
Training data taken without consent, biometric surveillance, employee monitoring, hidden data sharing, or invasive profiling.
Workers replaced or tracked
Automation plans, layoffs tied to AI systems, productivity scoring, call-center replacement, or undisclosed workplace monitoring.
Government or corporate overreach
Secret AI scoring, automated eligibility decisions, predictive policing, public-sector tools, or systems people cannot appeal.
Strong tips include
Evidence, Context, Stakes
A small detail can matter if it is specific and verifiable. You do not need a complete investigation before contacting us.
Who is involved: company, agency, vendor, project name, executives, contractors, or public officials.
What happened: describe the system, decision, risk, breach, deployment, or internal concern in plain language.
When and where: include dates, locations, meetings, emails, release timelines, jurisdictions, or affected facilities.
How you know: firsthand involvement, internal documents, screenshots, logs, public records, customer reports, or community evidence.
Why it matters: who is affected, what harm could happen, what has been hidden, and whether the issue is ongoing.
Why tips matter
Recent Stories Show the Pattern
AI harm often becomes visible first through insiders, affected workers, community members, researchers, customers, and people who notice a system before it becomes a headline.
AI infrastructure
The energy cost of AI is now a public policy issue
The IEA's Energy and AI report tracks how AI data centers affect electricity demand, energy security, affordability, and emissions.
International Energy Agency / April 2025
Surveillance
Facial recognition false matches can hurt real people
The FTC said Rite Aid's facial recognition program produced thousands of false-positive matches and led to shoppers being followed, searched, or accused.
Federal Trade Commission / December 2023
Election misuse
AI voice cloning moved from theory to enforcement
After AI-generated robocalls imitated President Biden in New Hampshire, the FCC ruled AI-generated voices in robocalls illegal under existing law.
Associated Press / February 2024
Cybersecurity
State-linked actors have used generative AI in cyber work
Microsoft and OpenAI reported disrupting adversaries that used generative AI for reconnaissance, phishing drafts, coding support, and target research.
Associated Press / February 2024
Data exposure
AI chat systems can leak sensitive operational data
Security researchers found an exposed DeepSeek database containing user prompts, logs, and API authentication tokens, according to WIRED.
WIRED / January 2025
Safety research
Experts are warning about misuse, failures, and systemic risk
The international report synthesized evidence on advanced AI risks, including malicious use, technical failures, labor disruption, and systemic impacts.
International AI Safety Report / January 2025
How we handle tips
We Review, Verify, and Corroborate
We look for documents, firsthand accounts, public records, technical signals, affected people, and patterns that can be checked. We may follow up for more context. We publish only when the facts are strong enough to support the story.
Important security note
Email is not an encrypted whistleblower dropbox. Do not break the law to obtain material, do not send anything that would put someone in immediate danger, and do not use a work device or account if that creates risk for you. For highly sensitive material, consider getting legal advice before sending.
ChatGPTAiML Tips