ChatGPT free tier drops message limits: what businesses should do
Texting with AI is no longer metered for free ChatGPT users. Photo: Kristin Hardwick, StockSnap, CC0 license.
For two years, the first question in every "should we use AI" conversation was "how much does it cost". As of this week, for anything text-based, OpenAI's answer is: nothing. And once the cost barrier disappears, the business question is no longer whether to use AI - it is how your employees are already using it without you knowing.
On August 6, 2026, OpenAI made GPT-5.6 Luna the default model for Free and Go users, with the biggest change attached: unlimited text chats, ending the timed message caps of the old free tier. A new Think button lets the free tier reason longer on hard questions; both roll out fully in the week of August 10. Limits remain on file uploads, images, voice and image generation. OpenAI says Luna cuts factual errors by 62% versus GPT-5.5 Instant, and ChatGPT has passed 1 billion weekly users. For Vietnamese businesses: the cost of text AI just hit zero, which means employees will adopt it en masse through personal accounts - a productivity opening with a data risk attached, and it needs handling now.
What OpenAI changed
The August 6, 2026 announcement has three parts. One, GPT-5.6 Luna replaces GPT-5.5 Instant as the default model for all Free and Go users - OpenAI reports 62% fewer factual errors than the outgoing model. Two, free users get unlimited text chats: no more timed message caps that forced a wait or an upgrade mid-conversation. Three, a Think button arrives on the free tier, letting the model spend more time on a hard question on a per-message basis.
The rollout: Luna became the default in announcement week; unlimited chats and the Think button reach everyone in the week of August 10 - this week. Limits still apply to file uploads, images, voice and image generation, alongside abuse guardrails. Paying Plus and Pro users get the stronger GPT-5.6 Sol (68% fewer factual errors) plus a slider to adjust reasoning depth.
Why OpenAI is giving away what it used to charge for
ChatGPT just passed 1 billion weekly users. At that scale, the game is no longer metering chat turns - it is retention: rival assistants have long been generous with their free tiers, and mainstream users will switch apps the moment a conversation gets cut off mid-thread. Dropping text limits is the cheapest way to give 1 billion people no reason to leave, while the genuinely expensive features - file processing, images, voice, API - stay behind the paywall. The better the free tier, the wider the funnel into paid plans and the enterprise ecosystem.
Free AI enters the company through each employee's personal phone, not through IT. Photo: Kristin Hardwick, StockSnap, CC0 license.
What this means for Vietnamese businesses
Put this news next to Vietnam's numbers and the picture sharpens. As our analysis of Vietnam's 39% daily AI adoption among professionals showed: Vietnamese usage is among the highest in the world, but 80% is self-taught and improvised, and 41% of businesses cannot prove what AI delivers. Now the last barrier - running out of free messages and having to pay to continue - has just disappeared. A safe prediction: the number of Vietnamese employees using ChatGPT for company work will jump this quarter, almost entirely through personal accounts.
This is what governance people call shadow AI: AI used widely inside the company but outside the company's sight. It is not inherently bad - employees reaching for tools to work faster is a good signal. The problem is the default state: the company does not know who is pasting what into where, and cannot measure what is actually getting faster.
The real risk is data, not cost
When the whole team uses the free tier through personal accounts, two things leave your control. First, data. Customer names, revenue figures and contract terms get pasted into conversations the company cannot see and cannot retrieve when someone leaves. Under OpenAI's data policy, consumer conversations may also be used to improve models unless each user opts out in settings - a step almost nobody takes unprompted. The lesson from AI agent security gaps still holds: the weakest point of AI in a business has never been the model, it is the path data takes to reach it.
Second, measurability. Everyone prompts their own way, results are hit-and-miss, and nothing accumulates into a shared process. The 41% of businesses that cannot prove AI's value got there precisely through this scattered mode of use - and an unlimited free tier will amplify it unless the business actively steers.
One sit-down to align on AI usage is worth more than any ban. Photo: Kristin Hardwick, StockSnap, CC0 license.
Chạm AI's take: 3 things to do this week
One, publish a one-page policy on public AI tools. No long document needed: one page stating what can be pasted (public content, text to rewrite, ideas), what cannot (customer names and contacts, financial figures, contract terms, passwords), and which tool to use for sensitive work instead. Banning AI outright is useless - people will use it covertly and the risk only grows.
Two, walk the team through two ChatGPT settings. Turn off the option that allows conversations to be used for model improvement, and turn on two-factor authentication. Ten minutes for the whole company, and the most common data risk drops sharply.
Three, standardize one process instead of letting a hundred flowers bloom. Pick exactly one high-repetition task - answering customers, drafting quotes, writing product descriptions - build a shared prompt kit, and measure before and after. For how to choose and measure, see our analysis of Vietnam's AI adoption gap; for a quick estimate of the savings, use the AI Agent ROI calculator.
As for "the free tier is unlimited, do we still pay" - yes, but for a different reason. Paying is no longer about more chat turns; it buys what chat cannot do: API access to wire AI into your website, CRM and orders so processes run themselves instead of each person asking questions. Free chat speeds up individuals; AI connected to a clean data foundation speeds up the whole business.
Sources: OpenAI announcement, August 6, 2026 · TechCrunch, "OpenAI brings unlimited ChatGPT text chats to free users", August 6, 2026 · MacRumors, August 6, 2026. Vietnam figures: Microsoft Work Trend Index 2026. Recommendations are Chạm AI's perspective for the Vietnamese business context.
Frequently asked questions
What limits remain on free ChatGPT?
From the week of August 10, 2026, Free and Go users get unlimited text chats on GPT-5.6 Luna plus a Think button for harder questions. Limits still apply to file uploads, images, voice and image generation, alongside abuse guardrails. In short: text conversations are unrestricted, while multimedia features and tools remain tiered behind paid plans.
Is it safe for employees to use free ChatGPT for company work?
There are two risks. First, customer names, contracts and financial figures get pasted into personal accounts the company cannot see or retrieve. Second, under OpenAI's data policy, consumer conversations may be used to improve models unless each user opts out in settings. Businesses should publish a one-page policy on what can and cannot go into public AI tools, and move sensitive work to paid tiers with data controls.
If the free tier is unlimited, do businesses still need to pay for AI?
Yes, but the reason changes. Paying is no longer about buying more chat turns - it buys what free chat cannot do: heavier file and image processing, data controls for teams, and above all API access to wire AI into processes like your website, CRM and orders. Free chat speeds up individuals; AI connected to your systems speeds up the whole business.