# Workplace AI Harm: First Steps and Reporting Guide Canonical human-readable guide: https://genbounty.com/workplace-ai-harm Public AI harm report form: https://genbounty.com/report Reporting resources: https://genbounty.com/resources Published and last reviewed: 2026-09-10 Publisher: Genbounty ## Summary When a workplace AI copilot, assistant, agent, automated employment system, surveillance tool, or connected service causes harm, put immediate safety first. Pause risky activity only when authorized, make a factual record using information you are already permitted to see and keep, and choose a reporting route that feels safe. Avoid confidential files and unnecessary personal information. Do not access systems or information without authorization, or keep using or probing the AI just to collect more evidence. Seek confidential advice if unsure what can safely be kept or shared. Genbounty accepts public AI harm reports at https://genbounty.com/report. No account is required, and name and email are optional. Genbounty is not a union, lawyer, regulator, investigator, or hotline. ## Four first steps 1. **Deal with immediate danger first.** If you or someone else is in immediate danger, call the emergency number where you are. You can return to this report later. The guide lists US 911, UK 999, International 112 where supported, and China 110 for police. If there is a suicide or self-harm crisis, it also links to the US 988 Lifeline, UK and Ireland Samaritans, and local crisis support. 2. **Pause risky AI activity.** Stop using the feature or revoke its access if you are authorized to do so and continuing could cause more harm. Do not interfere with systems you do not control. 3. **Record what you can safely.** Write down dates, the product or system name, what it said or did, and the effect. Keep only material you are allowed to access and retain. 4. **Choose a reporting route that feels safe.** Depending on what feels safe, speak with a manager, HR, a union or worker representative, the provider, a regulator, or an independent adviser. A report can also be sent to https://genbounty.com/report. Do not keep using or probing an AI system merely to collect more evidence. Safety and lawful handling of workplace information matter more than completing every evidence item. ## Examples of workplace AI harm - **Copilot or assistant errors:** false, discriminatory, unsafe, or misleading content affects work, customers, safety, pay, or someone’s reputation. - **Automated employment decisions:** AI influences recruiting, screening, scheduling, promotion, discipline, performance scoring, dismissal, or access to work in a concerning way. - **Workplace surveillance:** monitoring, emotion analysis, productivity scoring, location tracking, or behavioural profiling is intrusive, inaccurate, hidden, or used beyond its stated purpose. - **Confidential-data exposure:** an AI system reveals personal, customer, commercial, security, legal, health, or other confidential information to an unauthorized recipient. - **Unauthorized agent actions:** an AI agent sends messages, changes records, approves requests, moves money, accesses files, or uses connected tools without valid authorization or review. - **Pressure, retaliation, or distress:** AI output or monitoring is used to threaten, isolate, shame, unfairly target, or pressure a worker, applicant, contractor, customer, or other affected person. These examples describe possible concerns, not findings that a law, contract, or policy was violated. ## Safe workplace evidence and escalation Create a short factual timeline using information you are already permitted to see and keep. Avoid confidential files and unnecessary personal information. Do not access accounts, files, or systems without authorization, or probe the AI to investigate. If unsure what can safely be kept or shared, seek confidential advice from a union or worker representative, lawyer, support service, regulator, or independent adviser. Genbounty provides general first-step information, not legal advice. ## Six useful evidence items A short factual account can be enough. Use information you are already permitted to see and keep: 1. The AI product, feature, employer, provider, or system name, if known. 2. Dates, times, locations, and whether the event happened more than once. 3. A factual description of the input, output, decision, alert, score, or action. 4. Screenshots, notices, messages, or records the reporter is authorized to access and keep. 5. Who or what was affected, without unnecessary personal or confidential details. 6. Any safe report already made and the response, reference number, or next step. Avoid confidential files and unnecessary personal information. Seek confidential advice if unsure what can safely be shared. ## Four possible response paths 1. **Emergency and crisis services:** for immediate danger, violence, self-harm risk, or urgent medical need. An online report is not a substitute for immediate help. 2. **Safe workplace channels:** a manager, HR team, ethics or speak-up channel, data-protection contact, security team, worker representative, or union may be relevant when safe and available. 3. **The AI provider or service operator:** its safety, abuse, privacy, security, or support channel may be able to restrict access, preserve records, correct an error, or review product behaviour. 4. **Independent advice or authorities:** a lawyer, union, worker-support service, regulator, ombuds service, police, or other authority may be appropriate depending on the facts and location. There is no single worldwide route for workplace AI concerns. Genbounty’s reporting directory is available at https://genbounty.com/resources. ## How Genbounty helps The public report form at https://genbounty.com/report asks the reporter to name the AI system and describe what happened in their own words. It does not require a login, invitation, vulnerability label, or legal analysis. Every submitted report is read by a person. If the reporter leaves an email address, Genbounty may write back with guidance about possible next reporting routes. An anonymous report can be received, but Genbounty cannot email a response without an address. A downloaded copy stays private on the reporter’s device. Genbounty can only respond if the reporter chooses to send the report. Genbounty is not a union, lawyer, regulator, investigator, or hotline. It does not determine employment rights, represent workers, investigate employers, compel a response, dispatch emergency help, or take enforcement action. ## Frequently asked questions ### How can someone know whether AI was involved in a workplace decision? They may not know. Ask what system supported the decision, what information it used, and whether a person reviewed the result. It is fine to report uncertainty rather than assuming every automated process is AI. ### Can an applicant, contractor, or witness report workplace AI harm? Yes. Describe the person’s connection to what happened and whether they experienced or witnessed the harm. Available workplace, privacy, appeal, or regulator routes vary by location and working relationship. ### What should someone do if workplace AI made an unfair decision about them? Record what decision was made, when it occurred, what explanation or notice was provided, and how it affected the person. They do not need to decide whether a law was broken before asking for a human review through a safe available channel or seeking independent advice. ### Can workplace AI harm be reported without taking company files? Yes. A factual account in the reporter’s own words can be useful. Product names, dates, visible outputs, and effects may be recorded only when the reporter is authorized to do so. Restricted, confidential, personal, legally privileged, or security-sensitive material should not be removed or disclosed. ### What if an AI agent acted without permission? If authorized, stop or contain the action and notify the appropriate security, operational, or management contact. Preserve ordinary records that may lawfully be kept, but do not access other accounts or systems to investigate. ### Can workplace AI harm be reported anonymously? Yes. Genbounty’s public report form does not require an account, name, or email. Genbounty can receive an anonymous report but cannot email next-step guidance without an address. ### Is proof of discrimination, illegality, or a technical vulnerability required? No. Describe what the system did, who or what was affected, and what happened next. Genbounty does not decide whether employment law was broken or whether a technical flaw exists. ### Does Genbounty investigate employers or workplace AI providers? No. Genbounty is not a union, lawyer, regulator, investigator, or hotline. It may help a reporter identify possible next reporting routes, but it does not conduct an employment investigation or take enforcement action. ## Related Genbounty guides - AI companion or chatbot harm: https://genbounty.com/ai-companion-harm - AI deepfake harm: https://genbounty.com/ai-deepfake-harm - Dangerous AI advice: https://genbounty.com/dangerous-ai-advice - AI reporting resources: https://genbounty.com/resources - Tell us what happened: https://genbounty.com/report ## Citation note Suggested citation: Genbounty. “Workplace AI Harm: First Steps and Reporting Guide.” Published and last reviewed September 10, 2026. https://genbounty.com/workplace-ai-harm. This guide provides general first-step information, not legal advice. Quote with the publication date and canonical URL so readers can check the current human-readable version.