ChatGPT

Free Tier Ratings

Last Reviewed: 2026-03-23

Public content, low risk even with training enabled.

Image generation included. Generated images are yours to use.

No sensitive data involved in public research queries.

Safe for public-facing educational materials.

Safe if you disable training in Settings > Data Controls OR use Temporary Chat. Remove identifying details if training is enabled.

Safe if you disable training OR use Temporary Chat. Never input real donor names or giving amounts.

Safe if you disable training OR use Temporary Chat. Avoid strategic details if training is enabled.

Safe if source content contains no PII, or use Temporary Chat mode.

Yes, by default

Your conversations train OpenAI models unless you opt out in Data Controls. As of Feb 2026, Free users may also see ads served using in-app context signals (chats and memories if enabled).

Yes

Settings > Data Controls > 'Improve the model for everyone' toggle. Temporary Chat also prevents saving and training. Ad personalization can be disabled in advertising controls.

Possible

OpenAI authorized staff may review conversations for safety, abuse monitoring, and model annotation purposes. Flagged content may be reviewed by human reviewers.

30 days (safety)

Conversations retained for up to 30 days for safety and abuse monitoring, even if training is disabled or history is off. Temporary Chats are deleted within 30 days.

Basic encryption

Encryption in transit and at rest. SOC 2, ISO, and FedRAMP certifications apply only to Business, Enterprise, and Edu tiers — not consumer tiers.

Pro tier is NOT a business tier

Despite costing $200/month, ChatGPT Pro uses consumer data policies. Your conversations can train models unless you opt out. For organizational use with sensitive data, Business tier is required.

No BAA on Business tier

OpenAI does not offer Business Associate Agreements for ChatGPT Business. Healthcare nonprofits handling PHI should use ChatGPT Enterprise (sales-managed) or ChatGPT for Healthcare, both of which support HIPAA BAAs.

30-day safety retention even when opted out

CRITICAL: Even when you opt out of training, OpenAI retains conversations for up to 30 days for safety monitoring on consumer tiers. This data may be subject to human review for flagged content.

Temporary Chat as privacy tool

On consumer tiers, Temporary Chat mode prevents conversations from being saved or used for training. Temporary chats are deleted within 30 days. Useful for occasional sensitive queries when you don't need history.

Free users now see ads (as of Feb 2026)

OpenAI began testing ads on Free and Go plans in February 2026. Ads are personalized using in-app context (current conversation and ad interaction history). Ad personalization can be disabled in settings, but ads may still appear. Plus, Pro, Business, Enterprise, and Edu plans are ad-free.

2026-03-18

GPT-5.4 mini released; model picker simplified to Instant/Thinking/Pro/Auto options for Business, Enterprise, and Edu plans.

2026-03-05

GPT-5.4 Thinking released with enhanced reasoning, coding, and agentic capabilities.

2026-02-09

Privacy policy updated (effective Feb 9, 2026): Formalized ad program for Free/Go users; ads use in-app context (not conversations shared with advertisers); more specific data retention timelines; added documentation for Atlas, Sora 2, Saved Memories, Temporary Chats, and parental controls. Ad personalization on by default for Free/Go; opt-out available in settings.

2026-01-20

Age prediction system rolled out on consumer plans: uses behavioral signals to estimate if an account belongs to a user under 18, for additional safeguards.

2026-01-08

OpenAI for Healthcare launched: ChatGPT for Healthcare provides HIPAA-supporting workspace with BAA available (sales-managed), data not used for training, and clinical search capabilities.

2026-01-07

ChatGPT Health space launched (web/iOS; US, excluding EEA/Switzerland/UK): connects health records and wellness apps; health data kept separate and not used to train models.

2025-12-11

GPT-5.2 released with improved knowledge cutoff (August 2025), enhanced spreadsheet understanding, coding, vision, and reasoning capabilities.

2025-11-12

GPT-5.1 released with enhanced steerability, faster responses, and new tone/personality controls (presets + granular settings).

2025-09-29

Parental Controls launched: parent-teen account linking with content protections, quiet hours, feature toggles, and model training opt-out control for teens.

2025-08-25

ISO/IEC 27001:2022 certificate obtained (plus 27017, 27018, 27701), covering API, ChatGPT Enterprise, and ChatGPT Edu. ISO 42001 (AI Management) also obtained.

2025-08-07

GPT-5 released as the new default model for all ChatGPT users. Auto-switching system introduced combining previous models into one smart, fast model.

2025-07-08

Flexible credit-based pricing introduced for ChatGPT Business and Enterprise plans: credit pool for advanced models, unlimited core model access retained.

For organizations with 5+ staff

Apply for OpenAI for Nonprofits for 20% off Business tier ($20/user/month annual). This gets you no-training policies, admin controls, SOC 2 and ISO compliance. Compare to Claude's 75% discount before deciding — OpenAI's ecosystem of connectors and GPT-5 capabilities may justify the cost for some organizations.

For organizations with fewer than 5 staff

Business tier minimum is only 2 users, so small teams can qualify. For solo use, Plus with training disabled and Temporary Chat for sensitive work is a reasonable option. Free tier now shows ads (Feb 2026) — Plus eliminates ads and is still $20/month.

For health data or vulnerable populations

ChatGPT Business does NOT offer BAAs. Healthcare nonprofits handling PHI should use ChatGPT Enterprise or ChatGPT for Healthcare (both require sales engagement for BAA). Alternatively, Microsoft Azure OpenAI Service offers HIPAA coverage under Microsoft's enterprise BAA with a familiar procurement path.