OpenAI Updates Usage Policies: June 2026 Key Changes

On June 9, 2026, OpenAI published a sweeping update to its usage policies and developer terms. The changes cut free-tier message limits, reduced API free requests, raised GPT-5 mini pay-as-you-go prices, and added new prohibitions on automated decision-making. The announcement was posted on OpenAI’s usage policy page. The new rules took effect the same day for consumer limits, while API pricing changes followed on June 15, 2026. OpenAI said the update was necessary to manage infrastructure demand and align with enterprise compliance expectations. The policy document grew from 4,200 words to just under 5,700 words. Several changes were buried in footnotes and appendix tables.

The update hit nearly every tier. Free ChatGPT users saw their message allowance drop to 12 prompts per 3 hours, down from 20 per 5 hours. ChatGPT Plus subscribers kept the $20 monthly price but lost daily file upload capacity, dropping from 50 files to 30. API developers on the free tier had their requests cut from 100 to 50 per day and were restricted to three models. Pay-as-you-go API customers faced an 18% input price increase on GPT-5 mini. Enterprise accounts did not see API price changes but received new data retention rules. This was not a single isolated change. It was a coordinated policy adjustment across consumer and developer products, as covered in our ChatGPT pricing changes report.

Why now? OpenAI faced intensifying price competition from Google and Anthropic throughout May 2026. Google cut Gemini API prices in mid-May, and Anthropic ended its agent subsidy on June 15. OpenAI’s update appeared to protect margins while shifting free users toward lighter models. The free tier became a stricter funnel rather than a generous product. Stanford HAI’s AI Index noted in May 2026 that developer API costs rose for flagship models across every major provider for the first time in two years. That broader context matters. OpenAI was not acting alone. It was following a pattern we have tracked in the AI free tier shifts across major providers report. The new policy also contained unexpected language about autonomous agents, a sign that usage rules were evolving faster than model capabilities.

Enforcement started immediately. OpenAI said repeated policy violations could trigger 14-day suspensions, and the new automated decision rules required written approval for lending, hiring, housing, and healthcare. Some developers reported receiving warning emails on June 10 for medical triage tools that were previously allowed. The company also updated its usage dashboard so users could see exact limits instead of guessing. That transparency was welcome. But the reductions themselves were real and measurable. If you relied on the old free tier for daily work, the change forced a hard choice: upgrade to Plus, switch to a competitor, or use a lighter open model. Our AI free tier limits get tougher report examined why free users are carrying more of the cost burden in 2026.

How Do the Top Options Compare?

Table reflects the changes announced in OpenAI’s June 9, 2026 usage policy update and developer terms. API pricing changes took effect June 15, 2026. Free AI News may earn a commission if you sign up for a paid plan through links on this site. This does not affect our reporting.

1. ChatGPT Free , Casual users testing ChatGPT without paying

ChatGPT Free users lost more than they expected on June 9, 2026. The headline change was a cut from 20 messages per 5 hours to 12 messages per 3 hours. That sounds close until you do the math. The old limit allowed roughly 96 messages per day if a user stayed awake and timed every window. The new limit produces 96 messages in the same 24 hours only if the reset timing is perfect, but the actual daily ceiling is lower because the shorter window makes accidental message use more punishing. More importantly, GPT-5 mini access disappeared for free accounts overnight. Free users were pushed back to GPT-4o mini and two older models. OpenAI buried this change in a footnote on the usage policy page. The free tier also gained new restrictions on file uploads and image generation. For many users, this was not an update. It was a downgrade. The move fit a broader trend. Free tiers across the industry have been tightened repeatedly in 2026, as we reported in AI free tier limits get tougher. Google and Anthropic had already made similar adjustments. OpenAI did not want its free tier to become the default workspace for serious projects. The company wanted to push heavy users toward paid plans. The policy update also clarified that free accounts could be suspended after repeated misuse, though OpenAI did not define misuse precisely. That ambiguity frustrated users who felt the rules shifted without warning. The new usage dashboard helps track the 3-hour reset, but it cannot bring back GPT-5 mini access. For casual users who ask a few questions per day, the free tier still works. But anyone who used ChatGPT Free as a daily assistant had to change habits immediately. The 12-message cap runs out fast when you are editing text, analyzing a PDF, or generating code snippets. The 3-hour reset means a mistake at hour two leaves you waiting just 60 minutes, but the smaller allowance makes that wait happen more often. The old 5-hour window was more forgiving because it held more messages per cycle. The new design feels faster but smaller. That is the core change.

Key strengths:

• ✅ Free access to GPT-4o mini remains available without a credit card.
• ✅ The 3-hour reset is easier to remember than the old 5-hour cycle.
• ✅ New usage dashboard shows exact remaining messages before each reset.
• ✅ Image generation limits are now stated in plain language.
• ❌ Daily capacity dropped roughly 40% for heavy users.
• ❌ GPT-5 mini access disappeared entirely for free accounts.
• ❌ No grandfathering or grace period for existing free users.

Who it’s for: Casual users who ask a few questions per day and can tolerate harder limits.

2. ChatGPT Plus , Paying subscribers who need daily productivity

ChatGPT Plus kept its $20 monthly price in June 2026, but the package changed underneath. OpenAI reduced daily file uploads from 50 to 30 files. That was a 40% cut in one stroke. The company also reduced custom GPT tool calls by 25% without a headline announcement. Subscribers discovered the change when their automations began failing on June 10. The usage dashboard showed a new tool call counter that was not present in the previous interface. Users who relied on custom GPTs for bulk summarization, data extraction, or code review hit the cap by mid-afternoon. OpenAI later confirmed the change in a support update. The rationale was infrastructure efficiency, but no credit or discount was offered. The timing was not accidental. Google had just cut Google AI subscription prices in late May, and Anthropic was preparing its own credit pool overhaul. OpenAI could not raise the Plus price without risking churn, so it reduced included capacity instead. This quiet capacity cut matched a pattern we covered in the AI price wars Google cuts OpenAI considers as competition heats up report. Paying customers felt the squeeze without seeing a higher bill. For many, hidden limits are worse than price increases because they distort budgets and workflows. On the positive side, Plus users retained priority access during peak load. They also kept access to GPT-5, while free users were pushed back to GPT-4o mini. The file upload cut mainly affected power users who batch process documents. Custom GPT tool call reductions hit developers who built internal tools on top of ChatGPT. The policy update did not change the Plus subscription fee. But the effective cost per task went up. Users who upload 50 files per day now need a second day or an upgrade to Team. That is a real cost shift even if the invoice looks identical.

Key strengths:

• ✅ Subscription price remained $20 per month, avoiding a headline price hike.
• ✅ Priority access during outages and peak times remained unchanged.
• ✅ New usage dashboard tracks file uploads and tool calls in real time.
• ✅ GPT-5 access stayed exclusive to paid tiers.
• ❌ Daily file uploads fell from 50 to 30, a 40% reduction.
• ❌ Custom GPT tool calls dropped 25% with no advance email.
• ❌ No refund or credit for the lost daily capacity.

Who it’s for: Paying users who need GPT-5 priority access and can manage lower file and tool caps.

3. API Free Tier , Small developers prototyping with OpenAI models

Developers on OpenAI’s free API tier faced a hard cut on June 9, 2026. The daily request allowance dropped from 100 to 50 requests per day. The model list shrank to three: GPT-4o mini, a text embedding model, and a whisper audio model. GPT-5 mini, which had been available on a trial basis, was removed from the free tier. The change broke many educational and hobby projects overnight. Students who used the free tier for class demos received HTTP 429 errors before lunch. The new rate limit page made the quota explicit, but the previous 100-request cap had already been tight for small chat applications. A 50% cut forced developers to either reduce polling frequency or move to paid plans. This update was part of a wider reset in API free access across the industry. We documented similar moves in AI API free tiers limits 2026. OpenAI’s free tier now looks comparable to Google’s tightened Gemini API free tier and Anthropic’s console credit approach. The change did not affect paying API customers directly, but it narrowed the on-ramp for new developers. That matters because developer mindshare has become a key battleground. If a new developer starts with a competitor because OpenAI’s free tier is too small, that has long-term revenue consequences. OpenAI appeared willing to accept that risk in exchange for reduced infrastructure load. The new 50-request daily cap was also paired with stricter rate limits per minute. A single burst could consume the daily quota in seconds. OpenAI added clearer error messages and a countdown timer in the console. That was a small but meaningful improvement. Developers still had no way to roll over unused requests. The free tier remained useful for testing authentication and one-off prompts, but it could no longer support even a small internal tool. The policy update also clarified that API keys with no usage for 90 days could be deactivated. That cleanup rule went into effect on June 15, 2026.

Key strengths:

• ✅ Still provides three core models without requiring a payment method.
• ✅ Clear quota errors and a countdown timer replaced silent rate limiting.
• ✅ No credit card is required to create a free API key.
• ✅ 50 daily requests are enough for authentication tests and simple prompts.
• ❌ Daily request cap dropped to 50 per day, half of the previous allowance.
• ❌ GPT-5 mini trial access ended for all free-tier developers.
• ❌ Projects with no usage for 90 days may be automatically deactivated.

Who it’s for: Hobbyists and students who need a small number of API calls and can live inside 50 requests per day.

4. API Pay-As-You-Go , Production applications with variable traffic

Pay-as-you-go API customers received the clearest price increase in the June 2026 update. GPT-5 mini input pricing rose 18% from $0.356 to $0.42 per 1M input tokens. Output pricing stayed at $1.68 per 1M tokens. The batch discount fell from 50% to 40%, which raised off-peak processing costs for customers who used the asynchronous API. OpenAI announced the pricing change on June 9, with an effective date of June 15. The company said the increase reflected higher inference costs for GPT-5 mini’s expanded context window. Developers with fixed budgets had to either reduce requests, switch to a smaller model, or renegotiate volume discounts. Independent data backed up the pressure. Stanford HAI’s AI Index reported in May 2026 that API prices for flagship and mid-tier models rose across every major provider for the first time since early 2024. That context means OpenAI’s move was not a unique event. But the 18% jump on a model that many startups had standardized on was painful. A company spending $10,000 per month on GPT-5 mini input tokens saw that line item rise to $11,800, before accounting for the batch discount cut. For a bootstrapped startup, that extra $1,800 matters. We covered similar billing shocks in major AI API pricing model updates June 2026 Anthropic Google and more. Some developers found relief by switching to open-weight alternatives hosted on their own infrastructure. Others moved batch workloads to Google’s Gemini Flash, which undercut GPT-5 mini on per-token pricing after Google’s May cuts. OpenAI’s batch discount reduction made that switch more attractive. The new token estimator tool in the dashboard helped teams forecast costs, but it could not offset the higher rates. The update also changed how usage tiers are calculated, moving from calendar-month billing to a rolling 30-day window. That change meant some customers lost access to volume discounts mid-cycle because their trailing usage dipped below the threshold. The rolling window rule took effect on July 1, 2026.

Key strengths:

• ✅ All production models remain available without long-term contracts.
• ✅ Batch processing still receives a 40% discount, better than standard routing.
• ✅ New token estimator tool helps teams forecast invoice changes.
• ✅ Volume discounts remain available for high-usage accounts.
• ❌ GPT-5 mini input price rose 18% to $0.42 per 1M input tokens.
• ❌ Batch discount dropped from 50% to 40%, increasing off-peak costs.
• ❌ Usage tier calculation moved to a rolling 30-day window, cutting some volume discounts mid-cycle.

Who it’s for: Production teams that can optimize tokens or shift batch workloads to lower-cost models.

5. ChatGPT Enterprise , Organizations with compliance and data governance needs

ChatGPT Enterprise did not face a direct API price increase in June 2026, but the usage policy update added compliance burdens. Data retention rules tightened across all enterprise contracts. The default retention window for API logs dropped from 180 days to 90 days. Organizations that needed longer retention had to request an exception and sign an addendum. The new policy also introduced explicit prohibitions on automated lending, hiring, housing, and healthcare decisions without written approval from OpenAI. That change forced compliance teams to audit existing workflows. Some health tech startups discovered that their symptom-triage chatbots, previously allowed under general terms, now fell into the restricted category. OpenAI sent warning notices to affected enterprise customers on June 12. The new rules were not entirely negative. Shorter default retention aligned with data minimization principles and reduced legal exposure for customers in the EU. The enterprise dashboard added audit logs for automated decision-making workflows. That feature let administrators flag models and tools that required approval. OpenAI also published a clearer list of high-risk use cases. The policy update removed some ambiguity that had existed since the original 2024 terms. For larger organizations, having explicit categories was easier to defend in an audit. Still, the compliance cost was real. Legal teams had to review hundreds of pages before June 15. Some customers delayed new AI features because they could not get written approval in time. OpenAI’s support queue for enterprise policy questions doubled in the first week after the announcement. The change interacted with the broader competitive picture. Enterprise buyers were already evaluating Google and Anthropic after their own billing and policy shifts. OpenAI’s new restrictions gave those competitors an opening. The ChatGPT pricing changes report noted that enterprise sales cycles lengthened in Q2 2026 as buyers demanded more granular usage forecasts.

Key strengths:

• ✅ No API price increase for enterprise contracts in June 2026.
• ✅ Default API log retention dropped from 180 to 90 days, supporting data minimization.
• ✅ New audit logs show which workflows use restricted model features.
• ✅ Clearer high-risk use case list reduced some legal ambiguity.
• ❌ Compliance teams had to update policies before June 15 with short notice.
• ❌ Automated decision-making prohibitions added legal review steps.
• ❌ Enterprise support queue doubled, slowing approvals for new projects.

Who it’s for: Organizations with dedicated legal and compliance staff that need clear data retention rules and audit logs.

Frequently Asked Questions

What Should You Remember?

• Free tier cut: ChatGPT Free dropped to 12 messages per 3 hours, roughly 40% less daily capacity.
• Plus capacity cut: ChatGPT Plus kept $20 monthly but lost 40% of daily file uploads and 25% of custom GPT tool calls.
• API free tier halved: Free API requests fell from 100 to 50 per day and lost GPT-5 mini access.
• GPT-5 mini price hike: Pay-as-you-go input token price rose 18% to $0.42 per 1M tokens effective June 15.
• New prohibitions: Automated lending, hiring, housing, and healthcare decisions now require written OpenAI approval.
• Enforcement risk: Repeated violations can trigger 14-day suspensions or permanent bans.

Free AI News is an independent editorial publication. Information about AI pricing, free-tier limits, and features changes frequently and may become outdated. Always verify current details through the vendor’s official pages. Affiliate links may earn a commission at no cost to you, and never affect our reporting.