ChatGPT

OpenAI processes and stores ChatGPT conversation data primarily in the United States, on Microsoft Azure infrastructure. Consumer-tier accounts (Free, Go, Plus, Pro) have no documented EU data-residency option; keeping data within the EU/EEA, Switzerland, or the UK is currently offered only to ChatGPT Enterprise, Edu, and API customers. This places ordinary consumer data under US jurisdiction, including exposure to US government access requests.

OpenAI encrypts ChatGPT data at rest using AES-256 and in transit using TLS. There is no end-to-end encryption of conversations: OpenAI's systems can access message content server-side, which the company says is necessary for abuse and safety monitoring. Enterprise customers can use their own encryption keys for data at rest, but this does not extend to consumer accounts and does not amount to E2EE.

OpenAI's privacy policy discloses sharing data with vendors, service providers, corporate affiliates (including Microsoft), successors in a business transfer, and law enforcement when legally required. The US policy also allows sharing limited data with marketing partners, subject to an opt-out, which goes beyond processors needed to run the service. No public evidence of outright data sales was found, but the sharing scope exceeds a strictly necessary processor model.

OpenAI may use the Free, Plus, and Pro conversation sessions to help improve its models, unless the user turns off the option "Improve the model for everyone" in Data Controls. This opt-out option is easy to find and has no effect on the service. However, it is not retroactive; any conversations already used for training will still be part of previous training runs. For Business, Enterprise, Edu, and API usage, the opt-out is the default setting, and no training occurs unless it is explicitly enabled.

ChatGPT is proprietary. OpenAI's Terms of Use prohibit attempting to discover or reverse-engineer the source code, algorithms, or components of the Services. The GPT-4 Technical Report states the company withholds architecture, model size, training compute, and dataset details because of the competitive landscape and safety implications of large models. No source code or model weights for ChatGPT's models are published, and no outside party can inspect, audit, or fork them.

OpenAI's Trust Portal publicly lists current SOC 2 Type II, SOC 3, and ISO/IEC 27001:2022, 27017, 27018, 27701, and 42001 (AI management) certifications, with reports available to verified requesters. The published scope covers ChatGPT Enterprise, Edu, Business, and the API but does not explicitly confirm coverage of the free and Plus consumer tiers. This is a gap in disclosure, even though the underlying infrastructure is shared across products.

ChatGPT is a closed-source, commercially operated product with no public contribution process, documented governance model, or code of conduct for external participation. Development decisions are made entirely internally by OpenAI, consistent with a proprietary product rather than an open project.

ChatGPT's core paid product is a tiered monthly subscription: Plus at $20/month and Pro at $100-200/month, plus a Business plan for teams. Since February 2026, OpenAI has run in-chat advertising on the Free and low-cost Go ($8/month) tiers, with OpenAI's Chief Revenue Officer stating the company is now "in the advertising business." Subscriptions remain the primary structure, but the free tier increasingly monetizes attention and data.

The free tier gives limited access to ChatGPT: about 10 messages per 5 hours on the flagship model before falling back to a lighter model, with a smaller daily cap for extended "thinking" responses. Free-tier users also see advertising as of 2026. It is usable for light tasks but not comparable to the paid tiers.

ChatGPT's paid tiers are listed with clear prices and features on OpenAI's pricing page. ChatGPT has a 1.6/5 Trustpilot rating from nearly 2,800 reviews, with recurring complaints about billing transparency and unexpected charges. Much of the cancellation friction comes from ChatGPT being billed through three separate systems depending on the signup platform (Apple, Google, or OpenAI directly) rather than a single deliberate dark pattern. No formal regulatory action against OpenAI on this point was found.

OpenAI launched "ChatGPT Go," a reduced-price tier (around $8/month, first introduced in India at roughly ₹399) to make ChatGPT more affordable in emerging markets, and expanded it to 170+ countries through 2026. This is a regional price adjustment, but OpenAI has not published the purchasing-power methodology or a full country-by-country price list, so the policy is not clearly documented.

ChatGPT's account data export (Settings → Data Controls → Export Data) delivers a ZIP file containing a structured conversations.json with full chat history and metadata plus an HTML viewer. The export covers all account data at once in an open, machine-readable JSON format.

There is no self-hosting option for ChatGPT. The models behind it are proprietary and API-only. OpenAI's only self-hostable release (gpt-oss) is a separate, smaller open-weight model line that does not power ChatGPT. Users depend entirely on OpenAI's infrastructure to use the product.

ChatGPT requires an active internet connection for every query. There is no offline mode for generating responses. The mobile apps allow read-only access to previously loaded conversation history while offline, but no core functionality works without connectivity.

ChatGPT is available on the web and as native apps for iOS, Android, Windows, and macOS. There is no official Linux, BSD, or F-Droid build. Linux users are limited to the web interface. This is multi-platform availability but entirely within closed, proprietary app-store ecosystems.

No public accessibility statement, VPAT, or WCAG conformance claim was found for ChatGPT. Independent accessibility reviews report inconsistent screen-reader support, including elements missing from the tab order and difficulty navigating chat history with a keyboard, with no evidence of a coordinated remediation effort.

In October 2025, OpenAI completed a major restructuring: the original nonprofit (now OpenAI Foundation) retains a controlling equity stake in a new for-profit entity, OpenAI Group PBC, a public benefit corporation. Microsoft holds roughly 27% of OpenAI Group PBC, from over $13B invested since 2019 — a large minority stake and close commercial partnership, but not majority ownership or formal control by a Big Tech parent.

OpenAI has not been acquired in the last three years but underwent a fundamental structural change in October 2025, converting from a nonprofit-controlled entity into a for-profit public benefit corporation (OpenAI Group PBC), with Microsoft's economic stake rising to about 27%. This is a substantial internal restructuring rather than a change of external owner, so it falls short of an acquisition by either a non-Big-Tech or Big Tech company.

OpenAI publishes public Usage Policies prohibiting harmful uses (weapons development, harassment, circumventing safety mitigations), an Ads Policy, and safety commitments including a Preparedness Framework and Safety Advisory Group. These are checkable public commitments but sit alongside default-on training on user conversations and a rapid 2026 pivot into advertising, which sit uneasily next to OpenAI's public safety-first framing.

No dedicated environmental or sustainability report, data-center energy disclosure, or renewable-energy commitment from OpenAI was found. The only public information on ChatGPT's environmental footprint is an informal, third-party-cited per-query energy estimate referenced in OpenAI's own education materials, not a coherent company policy.

ChatGPT launched publicly on November 30, 2022, and has been continuously maintained and updated since, with regular model releases through 2026 (the GPT-5.x line). At the time of this evaluation, it has been actively maintained for roughly three and a half years — established but short of the five-year threshold for the top score.

ChatGPT reported roughly 900 million weekly active users as of February 2026, and third-party tracking put it past 1 billion monthly active users by June 2026 — reportedly the fastest app ever to reach that scale. OpenAI has shipped a steady cadence of model releases through 2026 (the GPT-5.3 through GPT-5.6 line), showing an unmistakably active, heavily used, and continuously developed product.