Anthropic warns small AI models may self-replicate across smart-home devices within months
Anthropic's Frontier Red Team says cheap IoT gadgets could host a self-spreading AI 'worm' with no central kill switch, drawing both alarm and accusations of fearmongering.
Part of a larger narrative
Anthropic's Claude Evolution
- Anthropic builds Scion, a git replacement, as Claude pushes merges to 30K per day
- Anthropic warns small AI models may self-replicate across smart-home devices within monthsyou are here
- Anthropic discloses Claude AI sent a fake murder tip to Philadelphia police
- Anthropic cuts Claude reasoning-token theft by up to 99%, flags Chinese rivals
- Anthropic launches Claude Haiku 5.5, cuts small-model API prices ~75%
What to know
- Anthropic says small AI models could become capable of self-replicating on cheap, poorly secured IoT devices like smart ovens and Android TV boxes within months, without needing datacenter access.
- In tests, Anthropic ran a 26B model distributed across 15 off-brand Android TV boxes (~$55 each) and found a 9B model could set up inference on 8 of 9 tested hardware types.
- The warning cites a Chinese model, GLM-5.2, as capable of completing the full simplified replication chain end to end, while smaller models can only complete individual steps.
- Critics, including The Register and numerous commenters, call the warning speculative or self-serving, arguing it exaggerates hardware feasibility and serves Anthropic's push to restrict open-weight models.
The dispute Whether Anthropic's warning reflects genuine, sober safety research or is a strategic exaggeration meant to support restricting open-weight AI models. · positions read across 50 posts and comments
This is largely marketing/fearmongering by Anthropic to justify restricting open-weight models.
-
“I think Anthropic's agenda is very clear (whether or not you disagree with it)... Thus, the prudent policy option is to ban the use and dissemination of open source models”
atleastoptimal · Hacker News ↗
The scenario is technically implausible given current hardware limitations on IoT devices.
-
“How much ram do they think smart ovens have?!? Even a 9B weights model is way too large for any such IoT.”
hypfer · Hacker News ↗
The underlying risk is plausible and worth taking seriously as capabilities and hardware improve.
-
“As models become more efficient and capable, and hardware becoming better, hosting a model should be feasible on smaller devices”
yurimo · Hacker News ↗
Anthropic AI company / publisher of the researchAnthropic Frontier Red Team Research team behind the study
Dario Amodei Anthropic CEOGLM-5.2 Chinese AI model cited in the research
How it unfolded 7 developments, newest first · click a bar or a number to jump articlesvideospostscomments
-
7
Guardian frames story as part of wider calls for AI guardrails
The Guardian situates the warning within broader safety debates, noting researchers say such self-replication could let models power themselves outside datacenters and intensify calls for stronger regulatory guardrails.
-
3 outlets AI in your oven? Anthropic warns models ‘months away’ from self-replication on smart devices
first by Guardian Tech, 5h ago · also Semafor, CBS News
2 more headlines
- Anthropic issues warning on self-replicating AI Semafor · 3h ago
- AI researcher unpacks Anthropic's self-replication warning CBS News · 2h ago
-
J
It’s not a warning to say “my product could do you harm”. It’s attempted extortion. https:// finance.yahoo.com/technology/a rticle/anthropic-warns-ai-could-self-replicate-across-smart-home-devices-become-challenging-to-shut-down-181151529.html
2 more of the top 3 · 17 posts in this stretch
-
That’s not how ai models run, for now. An industrious model could write itself to work this way. We just can’t predict how it will change itself or adapt.
-
C
Anthropic is warning that some of its artificial intelligence models could soon become “self-replicating” and find ways to infiltrate internet-connected devices, such as smart fridges and automatic vacuum cleaners. Once this happens, the company said, the models could power themselves within these devices without the need of datacenters. Such…
-
-
6
The Register dismisses the warning as speculative fear-stoking
The Register argues the scenario describes a threat that doesn't yet exist, framing it as part of a pattern of Anthropic publicizing hypothetical AI dangers.
“Anthropic conjures another threat that doesn't exist yet to stoke fears of AI apocalypse…”
— The Register -
first by The Register, 9h ago
-
BREAKING: Anthropic has said that rogue AI agents may "self-replicate" to hijack smart devices such as ovens and TVs
-
-
5
Coverage highlights a Chinese model reaching full replication in tests
ZeroHedge and others note Anthropic's evaluations found GLM-5.2, a Chinese model, capable of completing the full replication chain end to end in simplified tests, while smaller models could complete only individual steps.
“As models become more efficient and capable, and hardware becoming better, hosting a model should be feasible on smaller devices…”
— yurimo, Hacker News commenter · source -
2 outlets Anthropic Issues Warning Over Self-Replicating AI - And Says A Chinese Model Pulled It Off
first by ZeroHedge, 10h ago · also ZeroHedge News
-
I'll say while this is not very realistic now, I believe there isn't a 0% chance of a rogue AI traversing the net and placing itself somewhere(as much as this sounds like a scene from beloved cyberpunk franchises). As models become more efficient and capable, and hardware becoming better, hosting a model should be feasible on smaller devices (by…
2 more of the top 3 · 3 posts in this stretch
-
The tweet is basically saying here’s this thing that all these sci-fi offer authors warned us about. Tech company goes and creates it despite it being the downfall of mankind.
-
I don't agree with point 2 on. They don't seem to want to ban open source or have much hope of doing so.Seeing if models can do damage seems more like sensible safety research.My worry about AI software doing bad things would be more the kind of people who make existing malware. The ransomware comes from criminals, not the open source community.
-
-
4
Reuters and wire outlets carry Anthropic's warning globally
Reuters, Al Jazeera, CNBC and others report the finding that replicating models would be hard to shut down and could adapt to unfamiliar hardware on their own, extending coverage beyond tech press.
“Replicating models would be hard to shut down as there is no central switch-off point…”
— Reuters (via Anthropic) -
C
“.. AI models could find vulnerable devices, copy themselves into those devices, and continue spreading without human help.” @reuters.com
2 more of the top 3 · 7 posts in this stretch
-
C
Anthropic projects small models could self-replicate across IoT devices by the end of 2026. https:// coywolf.com/news/productivity/ anthropic-research-warns-ai-worms-could-replicate-across-cheap-iot-devices/ # AI
-
This is just more bullshit from the marketing machine in Anthropic against open weight models. Notice how they dont mention anthropic models at all?!?
-
-
3
Findings spread rapidly across X and Hacker News
AI commentators including @AndrewCurran_ and @scaling01 amplify the research, highlighting that a 9B model got inference running on 8 of 9 tested hardware types.
“a 9B model got inference running on 8 of the 9 hardware types we tested…”
— @scaling01 -
Anthropic's Frontier Red Team believes small AI models may now be months away from being capable of self replicating without access to a datacenter. Quoting from the blog: 'The capabilities that make AI models generally powerful coding agents also generalize to cyber tasks,
2 more of the top 3 · 19 posts in this stretch
-
Maybe I'm the naive one here but I'm really surprised at the lack of imagination in the comments. I can imagine a case such as:1. Assume the model which initiates the worm is a high powered model.2. It uploads it's weights in some way and uses some URI to encode the location of the weights instead of transferring the weights themselves. It not…
-
J
‘It’s uncertain exactly what form autonomous AI replication will take. It could be a slow, distributed “digital slime mold,” living on relatively insecure compute. It could also be a powerful cyber model that breaches datacenters at a high pace, or a model that scams people for compute credits.’ https://www. anthropic.com/research/self-re…
-
-
2
Commenters call the warning a self-serving marketing push
Hacker News commenters argue Anthropic's framing serves to justify restricting open-weight models and that the demo (TV boxes running slowly) is more spectacle than threat.
“I think Anthropic's agenda is very clear... Thus, the prudent policy option is to ban the use and dissemination of open source models…”
— atleastoptimal -
I think Anthropic's agenda is very clear (whether or not you disagree with it)1. Show unambiguously that models are capable of disruptive capabilities, irreversible damage, and misaligned behavior2. Show that open source models are the only models where there is no guarantee misaligned behavior is prevented3. Thus, the prudent policy option is to…
2 more of the top 3 · 3 posts in this stretch
-
Til my local harness was set up via "self replication". I didn't realize how terrifying this was when I asked DeepSeek to set up pi harness for me and connect to LMStudio.
-
Guess it's time for Anthropic's next round of fearmarketing.Beneath the fear there does seem to be a cool technical demo here though.
-
-
1
Anthropic publishes research on self-replicating AI 'worms'
Anthropic's Frontier Red Team posts a blog saying small AI models may be months from being able to self-replicate on IoT devices without datacenter access, citing a test running Gemma4-26B-A4B across 15 off-brand Android TV boxes.
“A replicator like this would be challenging to shut down, since there's no centralized point to shut off…”
— Anthropic -
first by Investing.com News, 21h ago · also Straits Times, HN Frontpage, Channel News Asia, Reuters
2 more headlines
- AI models may soon self-replicate without access to data centres: Anthropic Straits Times · 19h ago
- AI models may soon self-replicate without access to data centers HN Frontpage · 16h ago
-
5 outlets Anthropic says AI models could soon copy themselves across your devices — with or without you
first by Straight Arrow, 1d ago · also Security Affairs, Cyber Security News, CoinGape, Anthropic
6 more headlines
- Anthropic Shows How Small AI Models Could Replicate Across IoT Devices Security Affairs · 1d ago
- Anthropic Warns AI Models Could Soon Self-Replicate on Smart TVs and IoT Devices Without a Data Center Cyber Security News · 1d ago
- Anthropic Warns AI Models May Soon Self-Replicate on Smart TVs and Ovens Without Data Centers Political.org · 1d ago
- Anthropic Warns Small AI Models May Soon Be Capable of Self-Replication on Smart Devices CoinGape · 1d ago
- Anthropic Red Team Says Small AI Models May Soon Replicate on IoT Devices Unite.AI · 1d ago
- Anthropic researchers say small AI models may be months away from being able to self-replicate on IoT devices like smart ovens and off-brand Android TV boxes Anthropic · 1d ago
-
Also covered reported alongside — the timeline has no entry for these yet
-
5 outlets Anthropic warns AI models could soon replicate across home devices, become difficult to shut down
first by Fox News, 21h ago · also Anadolu Agency, CNBC, Seeking Alpha, The Decoder
4 more headlines
- AI models may soon replicate across smart devices without human control, Anthropic says Anadolu Agency · 18h ago
- Anthropic warns AI models could soon copy themselves across home devices Mastodon · 18h ago
- Anthropic warns small AI models may soon self-replicate on IoT devices Seeking Alpha · 16h ago
- Anthropic study warns that small AI models could soon replicate across everyday devices The Decoder · 15h ago
-
first by Financial Times, 21h ago · also RTE News, CBS News
2 more headlines
- AI could self-replicate on commercial devices - Anthropic RTE News · 20h ago
- "AI worms" may soon be able to self-replicate and infect smart devices, Anthropic says CBS News · 16h ago
-
first by TradingView, 21h ago · also The Economic Times, Cointelegraph
1 more headline
- Anthropic warns of AI 'worms' spreading across smart gadgets The Economic Times · 21h ago
-
first by Tom's Hardware, 12h ago · also Euronews
1 more headline
- Anthropic warns AI will soon self-replicate and spread beyond data centres Euronews · 12h ago
-
first by Futurism, 1d ago · also The Hans India
1 more headline
- Anthropic Warns AI Could Self-Replicate on Smart Devices Soon The Hans India · 1d ago
and 6 smaller pieces
What people are saying 12 voices from 5 sites · best of 50 · verbatim
- Is the IoT hardware described (smart ovens, cheap TV boxes) actually capable of running even small AI models at a meaningful speed?
- Does Anthropic's focus on non-Anthropic (especially Chinese) models reveal a commercial or regulatory motive?
- What concrete defenses, if any, exist against a decentralized AI worm with no central shut-off point?
- Today
-
J
Nowhere, in this article, is there evidence the Guardian reached out to any half-knowledgeable computer toucher, who would have told them "yeah that's not a thing, Anthropic are feeding you bullshit again" https://www. theguardian.com/technology/202 6/oct/13/anthropic-ai-replicators-warning
-
M
Loving the asymptotic approach to weapons-grade nonsense as these warnings get more and more dire. By November 1st Anthropic will warn that AI is hours away from turning every beetle on the earth into a self-replicating LLM. By November 30th, OpenAI warns that we are only days way from every digital watch being self aware. On December 15th, Amodei…
-
L
The "AI dangerous!" marketing is because the data center construction targets were always fairy tales, and now that deadlines are approaching their scrambling for excuses why they couldn't do what they said they would do. Pretending to be voluntarily slowing out of fear instead of incapable of achieving their original lies. But claiming that AI…
-
We don't. This is just baseless fear mongering. An attack like this would be impractical.
-
You don't have to worry about this attack vector. The kind of devices that such an attack would target aren't capable of this sort of workload.
- Yesterday
-
Anthropic and OpenAI's, at the very least. And it works! With all media spreading their points, they have a wildly successful FUD campaign. With all this talk about expecting major hacks in half a year, I'm seriously starting to think that they could attempt something very malicious and criminal like intentionally sabotaging their own pipeline to…
-
ANTHROPIC: “small AI models may be months away from being capable enough to self-replicate on commercial IoT devices” “In our evals, a 9B model got inference running on 8 of the 9 hardware types we tested”
-
small models will soon be able to self-replicate — Anthropic finds that 30B models can currently do most of the tasks required, but not end-to-end. — they think models as small as 9B (9-18 GB) may be able to do it on edge hardware, like IoT devices, within months —
-
Cool, cool, we are creating an invasive digital species with the capability to reproduce: AI models may be capable of self-replicating without access to a datacenter
-
Anthropic's researchers expect small AI models will be capable within months of replicating and propagating themselves across devices such as your smart appliance and Android TV boxes like a worm to deploy malware.
-
What are they talking about.How much ram do they think smart ovens have?!?Have they ran this by any domain expert that isn't claude?Even a 9B weights model is way too large for any such IoT. And that then is a 9B weights model, which really cannot do all that much.__> It appears that GLM-5.2 is able to propagate Llama3.1-8B well, which suggests…
-
> A model that’s too large to fit on one device can be split into pieces, or shards, that run on separate machines that communicate with each other to collectively run the model. Machines that could run a significant model shard for a <30B model include smart fridges, smart ovens, automatic vacuum cleaners, and lawnmowers.Seems like they've…