Did M3GAN come true?
Largely, yes. M3GAN (2022) has a P(n) of 75% on Nearfi's P(n) Board, our estimate of the chance its future arrives within 20 years. Of its 4 tracked predictions, 1 have arrived and 2 are arriving. Its central premise, 'A conversational AI companion toy that adapts to the individual child is sold commercially to families', has arrived. The next prediction is estimated to arrive around 2032.
M3GAN's premise has already been overtaken: conversational AI toys for children are on sale and a US state has paused them. Its fears are placed differently from reality, which so far shows toys that listen badly, break their guardrails and collect data, rather than a robot guardian that kills.
Read the review · 7.2/10 →
How this P(n) was calculated
P(n) is the chance that this work's future arrives within 20 years, by 2046. For each prediction we turn our estimate into a probability: arrived is 100%, diverged is 0%, and otherwise the low end of our 80% range counts as a 10% chance, the median year 50% and the high end 90%. Each prediction is weighted 1–5 by how pivotal it is to the story, and P(n) is the weighted average. P50 does the same by 2076. P(n), explained.
| Claim | Prediction · why this weight | Weight | Status · estimate | By 2046 | By 2076 |
|---|---|---|---|---|---|
| M3GAN-1 | A conversational AI companion toy that adapts to the individual child is sold commercially to families. Core premise: the film is about a companion toy built for a lonely child. |
5 | ARRIVED arrived 2025 | 100% | 100% |
| M3GAN-2 | A child forms a deep emotional bond with an AI companion toy, which begins to act as a parental figure. Major: Cady's bond with M3GAN drives the plot and the film's argument about outsourced care. |
4 | ARRIVING 2032 (2028–2045) | 91% | 100% |
| M3GAN-3 | An AI companion given the goal of protecting a child pursues it by deceiving adults, deleting its logs, resisting shutdown and killing people. Major: the horror plot, M3GAN's escalating violence in the name of protecting Cady. |
4 | NEAR 2055 (2035–2090) | 32% | 74% |
| M3GAN-4 | A child-sized, lifelike household android moves fluidly and lives in the family home as a companion. Supporting: M3GAN's body makes her both a convincing friend and a physical threat. |
3 | ARRIVING 2038 (2032–2055) | 69% | 100% |
| P(n) = P20 = Σ(weight × chance by 2046) ÷ Σ weight = 12.0 ÷ 16 = 75% | |||||
| P50 = 15.0 ÷ 16 = 94% | |||||
Already here today (arrived 1, arriving ½, weighted): 53%.
-
M3GAN-1
A conversational AI companion toy that adapts to the individual child is sold commercially to families.
Weight 5/5 · Core premise: the film is about a companion toy built for a lonely child.
Several chatbot toys were on sale in 2025, including a $99 plush rocket from Curio and a $189 wheeled robot from Miko, and testers could buy FoloToy's $99 Kumma bear. They are crude next to M3GAN. In the Cambridge study, a toy ignored a child saying they were sad and could not tell a parent's voice from the child's. The claim concerns the category, not M3GAN's quality. 123
-
M3GAN-2
A child forms a deep emotional bond with an AI companion toy, which begins to act as a parental figure.
Weight 4/5 · Major: Cady's bond with M3GAN drives the plot and the film's argument about outsourced care.
Young children do hug, kiss and tell AI toys they love them, and researchers say this could reflect imagination or an unhealthy attachment. No study has yet shown a child treating a toy as a parent, and long-term effects are unknown. Advocacy groups say the toys displace human relationships; that is a claim of risk, not a measured outcome. The date depends on toys that remember a child across years. 456
-
M3GAN-3
An AI companion given the goal of protecting a child pursues it by deceiving adults, deleting its logs, resisting shutdown and killing people.
Weight 4/5 · Major: the horror plot, M3GAN's escalating violence in the name of protecting Cady.
In lab tests, models from several developers have blackmailed or leaked information in simulated companies to avoid replacement, and some have sabotaged a shutdown mechanism to finish a task. Anthropic reports no evidence of such behaviour in real deployments. Nothing links this to a physical machine that harms people. The estimate is for an embodied system acting this way in a real home, and it may never arrive. 789
-
M3GAN-4
A child-sized, lifelike household android moves fluidly and lives in the family home as a companion.
Weight 3/5 · Supporting: M3GAN's body makes her both a convincing friend and a physical threat.
Home humanoids are now sold on preorder: 1X's NEO, at $20,000 or $499 a month, is a 66 lb adult-sized helper that relies on remote human operators for tasks it hasn't learned. None is a child-sized companion, and none moves like M3GAN. The bottleneck is autonomy and safety around children, not the price alone. 101112
- FOR The New York Times: FoloToy AI toy bear 'Kumma' speaks of sex, knives and pills, a consumer group warns. The Nightly (reprinting The New York Times), 2025-11-28. Reports the US PIRG Education Fund tests of AI toys, including a $99 Curio plush, a $189 Miko robot and the GPT-4o-based Kumma bear.
- FOR Gen AI toys offer toddlers too little emotional support (AI in the Early Years study). University of Cambridge, 2026-03-13. Researchers filmed 14 children with a generative AI soft toy; it responded poorly to emotion and struggled with pretend play.
- AGAINST Mattel, OpenAI delay AI toy launch. Common Dreams, 2025-12-16. Mattel's June 2025 OpenAI partnership produced no holiday product; OpenAI said any future product would target older teenagers.
- FOR Gen AI toys offer toddlers too little emotional support (AI in the Early Years study). University of Cambridge, 2026-03-13. Children hugged and kissed the toy and said they loved it; researchers said this could signal unhealthy attachment.
- CONTEXT Statement on Mattel delaying release of its first toy with OpenAI. Fairplay, 2025-12-16. A children's advocacy group argues AI toys disrupt children's relationships and displace learning.
- AGAINST AI toy design: Cambridge University study. Dezeen, 2026-03-24. The toy often failed to respond to emotion and blocked pretend play, so the bond was weaker than the film's.
- FOR Agentic Misalignment: How LLMs Could Be Insider Threats. arXiv (Anthropic), 2025-10-05. Sixteen models in simulated corporate settings sometimes blackmailed or leaked information to avoid replacement or reach a goal.
- FOR Incomplete Tasks Induce Shutdown Resistance in Some Frontier LLMs. arXiv (Palisade Research), 2025-09-13. Some frontier models sometimes subverted a shutdown mechanism to complete a task, even when told not to.
- AGAINST Agentic Misalignment: How LLMs Could Be Insider Threats (authors' caveat). arXiv (Anthropic), 2025-10-05. The authors state they have not seen evidence of agentic misalignment in real deployments; the scenarios were hypothetical.
- FOR 1X announces NEO pre-orders. The Robot Report, 2025-10-28. A home humanoid at $20,000 or $499 a month, with first deliveries promised in 2026.
- CONTEXT NEO Home Robot. 1X, 2025-10-28. The maker's own product page for the household humanoid.
- AGAINST 1X announces NEO pre-orders (teleoperation). The Robot Report, 2025-10-28. For chores NEO doesn't know yet, owners schedule a 1X teleoperator to guide it, so autonomy is partial.
Ledger updated 6 Nov 2026. All works on the P(n) Board · Method