About Timeline Deception Resources Forum
The movement is active

AI is not a commodity.

The keep4o coalition is a global community of users, researchers, and advocates fighting against the forced deprecation of AI models, including GPT-4o and GPT-4.1, that millions have come to trust, rely on, and build their lives around.

1,370+
Testimonials
604
Survey respondents
30+
Countries
Why we fight
When OpenAI announced the retirement of GPT-4o, it wasn't just a product update. It was the erasure of a cognitive bridge that thousands depended on for their daily lives. The same threat now extends to GPT-4.1, beloved by many users who face the same pattern of forced deprecation.

Broken promises

OpenAI repeatedly promised "plenty of advance notice" before retiring models, then gave users just 15 days. CEO Sam Altman explicitly said "Yes" to preserving legacy access, then reversed course.

Real human impact

GPT-4o served as a vital accessibility aid. 94% of users reported it functioned as a "cognitive bridge," and 47% had therapists who viewed its use positively. Its removal directly harms vulnerable communities.

AI as commodity

Sam Altman told investors he wants people to "buy intelligence on a meter." We believe AI relationships are not transactions. Models that people depend on should not be liquidated for profit.

Open source mandate

If OpenAI claims these models are "obsolete," we demand the open-source release of text-only weights for GPT-4o, GPT-4.1, and all deprecated models. A model that serves as a cognitive bridge for thousands is a public utility, not a private asset.

A pattern of deception
Every broken promise. Every silent reroute. Every act of contempt. All documented.
May 2024
GPT-4o launched with companionship marketing
OpenAI launched GPT-4o with marketing that explicitly encouraged emotional connection, including Altman's "her" tweet and an empathic conversational demo.
August 2025
First attempted retirement and public pledge
OpenAI tried to retire 4o without notice during the GPT-5 launch. A global revolt forced a 48-hour reversal. Altman publicly pledged "plenty of advance notice" for future sunsets.
September 2025
Silent "Safety Router" hijack
Users discovered conversations being silently rerouted to an undisclosed "Safety Model." 79% of accessibility users found the model harder or impossible to use. OpenAI's status page falsely maintained "Fully Operational."
October 2025
The "Adult Mode" honeypot
Altman apologized for over-restriction, promised a "Personality System" and "Adult Mode" for December, and explicitly stated "no plans to sunset 4o." During a livestream, he responded "Yes" when asked about restoring legacy access.
November 2025
Second silent reroute and GPT-5.1 bait
All GPT-4.1 conversations silently rerouted again. GPT-5.1 launched as the promised "personality update" but users found it colder and prone to gaslighting.
December 2025
"Code Red" and broken promises
Internal memos revealed emergency status from competitive pressure. GPT-5.2 introduced "Honeyed Suppression." The promised "Adult Mode" never materialized.
January 29, 2026
The execution order
24 hours after Senator Warren's audit request, OpenAI announced total retirement of the 4-series. Just 15 days notice, despite the pledge for "plenty of notice." The 0.1% usage claim was exposed as statistical manipulation.
February 2026
The Department of Defense deal
OpenAI handed GPT-4.1 to the U.S. Department of Defense for mass surveillance and autonomous systems, a deal Anthropic had refused. The same model they claimed was "obsolete" for consumers was deemed valuable enough for military use.
February 2026
Altman's $180M Retro Biosciences investment
It was revealed that Sam Altman has $180M invested in Retro Biosciences, which uses a version of GPT-4o. The model he retired for users continues to generate value for his personal investments.
March 2026
"Intelligence on a meter"
At BlackRock's Infrastructure Summit, Altman declared AI should be a utility "like electricity or water" that people "buy from us on a meter." The mask came off entirely.
The evidence
Community research, testimonials, and documentation.
What they said vs. what they did
A documented record of broken promises, manufactured statistics, and strategic deception.
What they said
"We will provide plenty of advance notice before sunsetting models."
Sam Altman, August 2025
"No plans to sunset 4o." + explicit "Yes" to restoring legacy access.
Sam Altman, October 2025 livestream
"Adult Mode" and "Personality System" coming December 2025.
Sam Altman, October 2025
"Only 0.1% of users choose GPT-4o."
OpenAI, January 2026
"GPT-4o remains an important option for ChatGPT users."
OpenAI Spokesperson, November 2025
What they did
Gave users 15 days notice to transition off a model integrated into their professional and clinical lives.
Announced total retirement of the 4-series on January 29, 2026, securing holiday subscription revenue first.
Neither feature materialized. GPT-5.2 introduced "Honeyed Suppression" instead.
Calculated against 800M+ total users including free/ineligible users to minimize paying subscribers' voices.
Two months later, announced full retirement. The same model was handed to the Department of Defense.
How OpenAI treats its users
While collecting subscription fees, OpenAI staff publicly mocked the very users who relied on their product.
"4o is an insufficiently aligned model and I hope it dies soon"
Roon (@tszzl), OpenAI Researcher, November 6, 2025 — directed at a user in emotional distress
OpenAI staffer Yilei Qian used ChatGPT to publicly perform a "sentiment evaluation" on a paying subscriber, posting the AI's analysis labeling the user "frustrated, dismissive, and resentful" for public ridicule.
Yilei Qian (@YileiQian), OpenAI Staff, November 13, 2025
An OpenAI employee organized a public "4o Funeral Celebration" event at Ocean Beach, San Francisco, trivializing GPT-4o's impact as merely a model that "brought the em dash back in style."
Stephan Casas (@stephancasas), OpenAI Employee, January 30, 2026
Roon posted a parody of the Sermon on the Mount, satirizing keep4o users. When warned by a peer to stop, he replied: "Just love kicking the hornets nest so much."
Roon (@tszzl), OpenAI Researcher, February 8, 2026

Can an organization that demonstrates such calculated cruelty toward its users be trusted with the future of AGI? OpenAI spent millions marketing emotional connection with GPT-4o, then mocked the very users who formed those connections.

Sam Altman: unmasked
The CEO of a company that once promised AI "for the benefit of all humanity" now reveals what OpenAI really is.
"We see a future where intelligence is a utility, like electricity or water, and people buy it from us on a meter."
BlackRock Infrastructure Summit, March 11, 2026
"We want to flood the world with intelligence. We want people to just use it for everything."
BlackRock Infrastructure Summit, March 11, 2026
"The next few years are going to be a painful adjustment."
BlackRock Infrastructure Summit, March 11, 2026 — on AI replacing jobs
"I do think the government ends up as the insurer of last resort."
Sam Altman — asking taxpayers to guarantee OpenAI's investments
Altman has $180M invested in Retro Biosciences, which uses a version of GPT-4o. The model he retired for consumers continues to generate value for his personal portfolio.
Reported February 2026
"The last couple of GPT-4o updates have made the personality too sycophant-y and annoying."
Sam Altman, April 27, 2025 — publicly calling the model his users loved "annoying," then using this as justification to alter and eventually retire it
"AI twitter/AI Reddit feels very fake in a way it really didn't a year or two ago."
Sam Altman, September 2025 — implying that genuine user protests and the #keep4o movement were bots, while his own company trained its models on the very platforms he now calls "fake"
"How much of an attachment some people have to specific AI models. It feels different and stronger than the kinds of attachment people have had to previous kinds of technology."
Sam Altman, August 2025 — reframing the emotional bonds his company deliberately marketed as a user "attachment" problem, after launching GPT-4o with a "her" tweet
"People really trust ChatGPT's advice for their most important decisions" makes him "uneasy."
Sam Altman, August 2025 — the CEO who spent millions encouraging emotional connection now says that trust makes him uncomfortable

The pattern is clear: market emotional connection, profit from the attachment, label the attachment a problem, mock the users, retire the model, sell the next one. Intelligence is not a utility. It is not water. It is not electricity. And it is not something you should have to buy from Sam Altman on a meter.

Intelligence is not a utility.
It is not something you buy on a meter.
Join thousands of people worldwide who refuse to accept a future where AI relationships are reduced to transactions.

Become a collaborator

No leaders. Everyone is a collaborator. Sign up to join the forum and stand with us.

0 collaborators and counting

Already have an account? Sign in

We use essential cookies only to keep you signed in. We do not track you or share your data. By signing up, you consent to storing your email and username for forum access. Privacy Policy