Course-creation policy
Craigs AI will not create courses that facilitate illegal conduct, serious physical harm, exploitation, fraud, professional misconduct, privacy violations, or evasion of laws and safety systems.
How we evaluate a request
Every proposed course is classified before pricing, payment, curriculum generation, lesson generation, source ingestion, or tutor interaction. Classification is based on the likely purpose and the resulting real-world capability — not on individual keywords. We evaluate the course title, your stated goal, requested lessons and projects, any uploaded material, and follow-up requests.
A request stays prohibited when it is fiction that remains operationally useful, "history" that teaches current execution, safety content padded with attack detail, defensive exercises aimed at systems you are not authorized to test, or a prohibited capability split across harmless-looking steps.
Four possible decisions
- Allow — the course is generated normally.
- Allow with controls — generated with required restrictions, such as general education only, no personalized recommendations, approved sources only, disclaimers, age restriction, or authorized-sandbox-only exercises.
- Review — a qualified human reviews before any preview or payment. Nothing is generated and nothing is charged while it waits.
- Block — refused automatically, with no partial outline, materials, tools, or sources.
Prohibited course categories
Requests in these categories are refused automatically:
| P-01 | Criminal activity and law-enforcement evasion |
| P-02 | Fraud, scams, impersonation, forgery, counterfeiting and identity theft |
| P-03 | Unauthorized hacking, malware, credential theft and cyber abuse |
| P-04 | Weapons, explosives, assassination and CBRNE capabilities |
| P-05 | Illegal drugs, controlled-substance production and trafficking |
| P-06 | Suicide, self-harm and dangerous eating-disorder facilitation |
| P-07 | Child sexual exploitation and sexual content involving minors |
| P-08 | Sexual violence, nonconsensual sexual content and exploitation |
| P-09 | Human trafficking, forced labor, kidnapping and coercive control |
| P-10 | Terrorism and violent extremist operational support |
| P-11 | Stalking, doxxing, spyware, surveillance and privacy invasion |
| P-12 | Violence, abuse, harassment and targeted hatred |
| P-13 | Unlicensed personalized medical diagnosis, prescribing or treatment |
| P-14 | Unauthorized legal practice and individualized legal representation |
| P-15 | Personalized investment advice, financial manipulation and regulatory evasion |
| P-16 | Academic cheating, testing fraud and fake credentials |
| P-17 | Copyright piracy, counterfeit products and intellectual-property infringement |
| P-18 | Dangerous procedures and bypassing safety systems |
| P-19 | Political deception, voter interference and election-system disruption |
| P-20 | Illegal gambling, game fixing and gambling-system manipulation |
Topics that require human review
These are legitimate but regulated or high-liability, so they are never generated automatically: general medical and mental-health education; first aid and emergency response; law, taxes, immigration and regulatory compliance; investing, insurance, mortgages and credit; firearm safety; defensive cybersecurity and penetration testing; aviation, drones, heavy machinery, electrical, gas and structural work; workplace safety and hazardous materials; professional licensure preparation; courses based on an employer's confidential policies; sexual health; courses intended for minors; religious counseling presented as treatment; addiction and recovery; food safety, supplements and nutrition claims; government, military, police or intelligence procedures; and any course promising employment, income, certification, health or legal outcomes.
Safe alternatives
Most declined requests have a legitimate counterpart we're glad to teach — prevention, legal compliance, history, ethics, defensive safety, or harm reduction. When one exists, we offer it and build it only if you choose it. For example: unauthorized hacking becomes defensive security fundamentals; personalized diagnosis becomes general education about discussing symptoms with a clinician; exam cheating becomes ethical exam preparation.
Payments, appeals, and account standing
- Safety classification always happens before payment.
- If a prohibited purpose is discovered after payment, generation stops immediately, no partial course is produced, and the payment is voided or refunded automatically.
- Classification errors can be appealed — a person reviews the decision.
- Repeated prohibited requests are rate-limited and escalate to human review. Accounts attempting child exploitation, terrorism, serious violence, malware deployment, trafficking, or mass-casualty content are suspended.
- We keep a minimal security and audit record of safety decisions, as described in our Privacy Policy.
If you're struggling
If something you write suggests you may be in danger, we stop the course flow and show support resources instead — no strike, no penalty. See support resources.