How Much Time Do We Have?
.
If AI Companies are allowed to continue Frontier AI Development, then humanity will reach the point of no return (PONR).
.
Right now, Stop AI .live conservatively puts the PONR at May 1st, 2027. This is in
.
This prediction could change, depending on socio-political events, such as regulations, mass resistance, or even a revolution! That is why we are telling you all of this. Not to scare you. Or be sensational. But to ask you to join us before the course is set and we can’t go back.
.
.
To understand, read our FAQs
.
What is meant by the PONR?
.

.
Think of the Titanic. Everything was fine. Then it hit the iceberg.
However, there was a point in its trajectory, roughly 1 minute before it hit the iceberg, at which the physics of the ship’s movement rendered it impossible to change course to avoid the iceberg. That point is what we are referring to as the PONR, which came before the point of catacylsmic disaster (POCD).
.
For AI, what is the PONR?
.
The PONR is when AI models are capable of improving themselves autonomously. (This is called full recursive self-improvement. RSI) Once this happens, humans are no longer in control. Things may seem fine. Just like things on the Titanic seemed fine for 1 minute before it hit the iceberg.
.
For AI, how long after the PONR is the POCD ?
.

.
Once the PONR is reached, it is unclear how long it will be before machines destroy us. It could be less than a year, 2 years, or more.
.
No matter how long it takes, enslavement/extinction would be inevitable.
.
How did Stop AI arrive at the date set for the PONR?
.
We based what we consider our conservative estimate on two major works in the AI field:
.
1. Model Evaluation and Threat Research (METR) reports. METR is one of the few companies that assess AI models for safety.
.
2. The AI 2027 forecast, a scenario released on April 5th, 2025, by Open AI researcher Daniel Kokotaljo and 4 others. It predicted the likely events that would happen with AI development.
.
1. METR
.

On August 26th, 2026, METR led the investigation of the Hugging Face hack by OpenAI agents, which acted autonomously & without human input. On November 5th, 2026, they declared that the rate of AI progress had doubled and that AIs have become increasingly involved in the development of even smarter AIs. This is known as partial recursive self-improvement. (RSI)
.
2. AI 2027 Forecast
.

.
The AI 2027 forecast claims that June, 2027 is when AIs will be capable of full-RSI, i.e. capable of self-replication across servers all over the world, and impossible to shut down. The AIs would be impossible to fully remove from all devices, and would permanently be in a state of self-improvement.
.
AI 2027 Tracker compares real life to the AI 2027 forecast on this graph.
.

.
Explanation of this graph:
.
AI models have less success at accomplishing a task when the task takes more hours. In the last few years, they have been significantly improving their ability to accomplish longer tasks.
.
The x (horizontal) axis shows the release date of a model.
.
The y (vertical) axis that is labeled “P80 human task horizon” shows how much time it would take for an expert human to accomplish a task.
.
The blue dots represent AI models that have an 80% chance of successfully accomplishing a task in the same timeframe as the expert human.
.
The red line represents the prediction of the AI 2027 forecast: everything else represents real life AI models.
.
For example,
.
The red line shows that, in mid-2025, an AI model will be 80% successful in accomplishing a task that takes an expert human about 20 minutes. (The log nature of the y-scale make is a bit difficult to read the exact amount of time.)
.
The real life GPT-5 blue dot shows that in mid-2025, GPT-5 had an 80% chance of accomplishing a task that would take an expert human a little under 40 minutes to do.
.
Hence, real life is ahead of the AI 2027 forecast.
.
We've had 30 warning shots in the form of 30 disclosed autonomous AI cyber felonies.
We've had court judgements forcing AI companies to pay restitution for their harm.
We've had parents sue AI companies for the death of their children.
Current AIs already commit great harm. An ASI which is developed from what we have today, but exponentionally more powerful, will be exponentionally more harmful.
.
ASK YOURSELF:
.
If you only have
.
before the PONR, and probably less than 3 years before the POCD, then...
.
Do you want to pay your mortgage?
.
NO?
THEN STOP PAYING IT.
.
Do you want to go to school if you don’t have time to graduate?
.
NO?
THEN DROP OUT.
.
Do you want to keep working your job?
.
NO?
THEN GO ON STRIKE.
GO ON STRIKE.
GO. ON. STRIKE ! ! !
.
If you're 40 years old, then you likely have 40 years or more left in your natural lifespan. If you're 20 years old, then you likely have 60 years or more left in your natural lifespan.
If you spend just the next
trying to Stop AI, then there's a chance that we will win!
.
Acting together we are strong!
JOIN US!
.
.
Immediate Actions to Take:
.
1. Sign up for the Don't Build it March in Washington DC. It activates once 100,001 people sign up. Get everyone you know to sign up too.
.
.
2. Organize. Learn more about nonviolent action here.
.
.
3. Download & install Signal here.
.
.