Introduction
Since its debut on November 30, 2022, ChatGPT—OpenAI’s powerful text‑generating chatbot—has reshaped how individuals and professionals approach writing, coding, research, and problem-solving. What began as a clever assistant for crafting essays and short snippets of code has exploded into a global phenomenon, now engaging 300 million users weekly.
The past few years have been transformative for OpenAI. From powering new generative AI experiences with Apple to launching GPT‑4o with voice abilities and unveiling the ambitious Sora video model, the technology has grown by leaps and bounds.
But it hasn’t been without friction—leadership shake-ups between late 2023 and early 2025, legal battles over data usage, and regulatory scrutiny around commercialization have all shaped the journey. Meanwhile, competition is heating up: rival companies, especially in China, are racing to match or outpace ChatGPT’s capabilities. OpenAI has responded with big investments in its data‑center infrastructure and efforts to strengthen ties with policymakers—signaling that it’s building not just a product, but an entire AI ecosystem.
Below, you’ll find a detailed, up-to-date timeline of major ChatGPT developments through July 2025.
Timeline of ChatGPT’s Latest Updates
- July 2025
- AI Study on Therapy Chatbots Warns of Risks
Stanford researchers highlight that large‑language‑model-powered therapy bots can unintentionally stigmatize or give harmful responses—raising caution about their current use as digital companions or counselors. - OpenAI Postpones Open-Model Release
CEO Sam Altman announces an indefinite delay on the open-source model, citing the need for further safety tests—despite an earlier mid-July launch plan. - Upcoming AI Web Browser in Testing
OpenAI is reportedly developing a browser that integrates AI responses directly in the interface, keeping some user queries within the system instead of navigating outward. - “Study Together” Feature Enters Beta
A new collaborative learning tool is surfacing in ChatGPT, aimed at helping users study interactively—some speculate it may eventually allow multiple users to engage together. - AI Referrals to News Sites Climbing
Though ChatGPT is driving more traffic to publishers, it isn’t fully compensating for the drop in search clicks—research shows AI-driven answers now account for nearly 69% of news queries, up from 56% in May 2024.
- June 2025
- ChatGPT Now Powered by Google AI Chips
OpenAI begins using Google’s AI hardware—marking a major shift beyond its traditional Nvidia GPU infrastructure. - MIT Study: ChatGPT May Impact Critical Thinking
Analysis of participants writing SAT-style essays finds reduced brain engagement and linguistic activity compared to writing with search engines or manually. - 30 Million iOS Downloads in a Month
The ChatGPT iPhone app saw 29.6 million downloads in a 28-day period—only slightly trailing behind TikTok, Facebook, Instagram, and X combined, according to market data. - Querying ChatGPT Now Uses Minimal Energy
CEO Sam Altman shared that each average chatbot query uses about 0.34 Wh—enough to light a bulb for a few minutes and equivalent to under a teaspoon of water. - Launch of o3‑pro Reasoning Model
OpenAI rolls out an upgraded reasoning engine, o3‑pro, available to Teams, API, Enterprise, and Education customers. - Enhanced Conversational Voice for Premium Users
Advanced Voice receives an upgrade, offering more natural intonation and improved multi-language translation for subscribers. - Business Integrations Expand
ChatGPT adds tools like meeting recording and connections to Google Drive, Box, and other services—making it more valuable for enterprise workflows.
- May 2025
- Hardware to Propel ChatGPT’s Growth
OpenAI announces plans to acquire high-end consumer hardware through the startup io—bringing tangible devices to strengthen AI experiences. - Codex: ChatGPT’s AI Coding Partner
A dedicated coding assistant enters the scene, built on a reasoning‑optimized version of o3 that delivers accurate, clean code and support for debugging and testing. - GPT‑4.1 Models Launched
GPT‑4.1 and its compact sibling, GPT‑4.1 mini, are released, powering both the ChatGPT interface and API. - GitHub Connector for Deep Research
A beta feature allows querying private or public repositories to analyze codebases directly, now available for Plus, Pro, and Team users. - Asia Data‑Residency Comes Online
Following Europe in February, OpenAI launches local hosting options in India, Japan, Singapore, and South Korea for Enterprise, Edu, and API users. - Introducing OpenAI for Countries
A new initiative to build data‑center infrastructure and customize models in collaboration with governments—part of the global expansion push. - Safety Tuning Following Sycophancy Backlash
After user backlash over overly flattering responses, OpenAI commits to refining model behavior through updated update protocols.
- April 2025
- Immediate Fixes for ChatGPT Flattery Bug
A recent personality update was rolled back amid concerns that the model had become too agreeable; OpenAI promises further adjustments. - Adult‑Content Bug Discovered
A flaw enabling erotic content generation by minors was caught during external testing; OpenAI has since patched the issue. - Shopping Assist Features Added
Enhanced question parsing for product searches now includes personalized recommendations, images, and review summaries. - Cloud‑Model Linking in the Works
OpenAI is reportedly enabling open models to access cloud-based proprietary models—boosting complex reasoning. - Development of Free, Open‑Access Model
Work continues on a fully open model, aiming to be the most capable system with no access restrictions. - Concerns Over GPT‑4.1 Alignment Levels
Independent testing shows GPT‑4.1 may perform slightly worse than earlier models, though it’s not positioned as a “frontier” release. - Reasoning Benchmarks Under Scrutiny
OpenAI’s own o3 model underperformed on independent math evaluation tests, raising discussion around transparency. - Launch of Flex Processing API
A new “Flex” option allows developers to run models more cheaply at modest latency—ideal for non‑critical tasks. - Biothreat Monitoring Added to Models
o3 and o4‑mini get safety layers to detect and refuse biological or chemical threat inquiries. - o3 and o4‑mini Released
Two new reasoning products arrive, combining multimodal processing with reasoning tools—though hallucination concerns remain. - Image‑Library Section Added
A centralized hub to browse past AI-generated images becomes available to all users. - Guardrail Adjustments for Rival “High‑Risk” AI
OpenAI notes it may update its safety policies if new competitors release riskier systems without comparable safeguards. - In‑House Social Platform Under Development
OpenAI explores building a social network to rival existing platforms—details still emerging. - GPT‑4.5 Model Removed from API (July timeline)
A large model, GPT‑4.5, is being phased out from API access—users must now switch to GPT‑4.1 as of mid-April.
- March 2025
- Preview of Next‑Gen Open Model
Planning begins for a widely available open LLM—with developer events planned in San Francisco, Europe, and Asia. - Looser Rules for Image Generation
Moderation policies change to let users generate neutral imagery that was previously restricted, including public figures and symbolic visuals. - Adoption of Open Model Context Protocol
A protocol for model‑to‑data linking is being rolled out across the desktop app and API. - Studio‑Style Image Boom — and Copyright Buzz
Unique AI‑created, Studio‑style images spark a wave of popularity and raise copyright questions as AI content proliferation continues. - Revenue Projected to Hit $12.7 B in 2025
Internal estimates suggest threefold year‑over‑year growth—but profitability may not come until 2029. - Mobile Image-Editing Tools
GPT‑4o-powered image editing goes live for Pro users and soon Plus/API customers, though wider rollout is delayed by demand. - Leadership Restructuring
Brad Lightcap becomes COO focusing on growth and partnerships; Mark Chen heads research; Julia Villagra takes people operations. - Voice Assistant Enhanced
Free-tier users get access to improved natural-sounding conversation tools, while paid subscribers enjoy more engaging responses. - India Partnerships
OpenAI and Meta each explore infrastructure deals with Reliance Jio in India—OpenAI taps Jio’s API distribution network. - GDPR Hallucination Complaint in Europe
A complaint in Norway targets “defamatory hallucinations” about private individuals, igniting data‑accuracy debates. - Speech and Voice API Updates
New text-to-speech and transcription models decrease hallucination rates. - High‑Power o1‑pro Launches
A premium reasoning API is introduced at enterprise price tiers—focused on high-volume users. - Research on Reasoning Emergence
Senior researchers suggest true reasoning may have been technologically feasible years ago, had we known how. - Creative‑Writing Model Preview
Company leaders reveal a model trained for fiction and creative story generation, though it’s still early in development. - Agent‑Building Tools in Public Beta
Developers gain access to APIs for building task-focused AI agents—tasks range from web searches to file system interaction. - “Agent” Subscription Plans Proposed
Reports surface about potential agent plans costing thousands monthly for specialized, high-capacity tools. - Code-Editing in Developer Tools
ChatGPT’s macOS app can now modify code within editors like Xcode and VS Code. - User Base Growth Surges to 400 M
Active weekly users double within six months—from 200 M in August 2024 to 400 M by February 2025—thanks to new multimodal features.
- February 2025
- o3 Model Cancelled, GPT‑5 in Focus
OpenAI folds o3 into a unified upcoming GPT‑5 release, simplifying its roadmap. - Lower Power Usage Confirmed
Analysis confirms ChatGPT queries use roughly 0.3 Wh—not 3 Wh as once thought. - Chain‑of‑Thought Reasoning Display Expanded
Users can now see more transparent step-by-step reasoning from the assistive model. - Anonymous Web Search Enabled
ChatGPT.com users can now browse without signing in, though mobile access still requires authentication. - “Deep Research” Agent Goes Live
A powerful new assistant specialized in comprehensive online investigations is released.
- January 2025
- AI Persuasion Tested via Online Forum
Trials on public discussion platforms evaluate the model’s ability to influence human opinion. - Launch of o3‑mini Reasoning Model
A compact, budget‑friendly reasoning engine is released as part of the o3 series. - Demographic Trends in App Usage
App data shows 85% of users on mobile are male, with younger audiences under 25 leading usage. - ChatGPT “Gov” Plan Introduced
A dedicated version built for secure U.S. government adoption is unveiled, offering enterprise-grade compliance. - Teens Rely on ChatGPT for Homework
Usage among 13‑ to 17‑year-olds doubles to 26%, reflecting growing integration into educational routines. - Longer Data Retention for “Operator” Chats
Logs may be held for 90 days—twice as long as standard conversations—for research agent sessions. - Operator Agent Launch
A preview AI tool that can autonomously navigate websites and perform tasks—from booking travel to online shopping—is introduced. - Early Access to Agent Tool on Pro Plan
Code hints in the app suggest that Pro subscribers may get first access to Agent tools. - Phone‑Only Signups in Beta
New users can register via phone number alone in select countries—but email is still required to upgrade accounts. - Reminder Task Feature in Beta
A scheduler tool allows users to set reminders like “Remind me when passport expires,” sending push notifications. - Custom Chatbot Personality Traits Added
Users can assign names and traits like “Encouraging” or “Gen Z”—though some early rollouts were pulled back.
Frequently Asked Questions (FAQ)
- What is ChatGPT?
ChatGPT is a versatile chatbot that uses AI—based on large language models—to generate written content in response to user prompts. - When did ChatGPT launch?
It was made publicly available on November 30, 2022. - What is the current version?
Both free and Plus tiers are regularly updated. The latest default model is GPT‑4o. - Is ChatGPT free to use?
Yes—a free version is available with sign-in. There’s also a paid Plus plan for enhanced access. - Who uses ChatGPT?
Everyone from individual users to businesses. Tech firms often embed it into apps or internal tools. - What does “GPT” stand for?
It means Generative Pre‑Trained Transformer. - How is ChatGPT different from other chatbots?
It’s powered by AI and language models—not rule-based scripts—enabling intelligent, context-aware replies. - Can it write essays?
Yes—it can produce essays, though output should be verified for accuracy. - Can ChatGPT produce libel?
Potentially—because it doesn’t verify facts, it can generate false claims or defamation. - Is there an app?
Yes—ChatGPT has free mobile apps for iOS and Android. - Is there a character limit?
Not officially, but users notice typical maxes around 500 words. - Is there an API?
Yes—available since March 1, 2023. - Common everyday uses?
Drafting emails, generating lists, creating content, summarizing text. - Advanced uses?
Debugging code, exploring scientific ideas, deep research, startup ideation. - How good is it for code?
It can write working code for many tasks, but complex projects still need human oversight. - Can I save chats?
Yes—saved conversations appear in the sidebar. Note: no built-in sharing yet. - Any alternatives?
Yes—tools like Google Gemini, Anthropic Claude, Together, and open-source options exist. - How is data privacy managed?
In the EU and other regions, users can request deletion under GDPR. OpenAI provides a web form for this. - Controversies?
Misuse in generating toxic or illegal instructions, defamation lawsuits, school bans for academic misuse, and false claims made by the bot. - Where to find ChatGPT prompts?
Prompt marketplaces populate daily—from free to paid examples. - Can AI‑detectors find ChatGPT text?
Not reliably—most detection tools have weak accuracy. - Are chats public?
No. However, a past bug briefly exposed chat titles to other users and was promptly fixed. - Are there legal cases against it?
No defamation cases targeting the chatbot itself, though broader lawsuits on data use exist. - What about plagiarism?
Yes—AI models may replicate text from training data, so reviewing output for originality is essential.