A Seller Claims to Offer Leaked Claude Fable 5 Weights From an AWS Bucket, a Claim No One Has Verified
馃毃馃嚭馃嚫 Anthropic model weights allegedly exposed through misconfigured AWS S3 bucket An actor using the handle "localizer" claims to have obtained a 7 GB shard of Anthropic's alleged "Claude Fable 5" model weights from a misconfigured AWS S3 bucket associated with Amazon Bedrock. The actor claims the complete model consists of 640 shards totaling approximately 8 TB, but says only one shard was retrieved before access was restricted. Screenshots showing a .safetensors file, model metadata, and a hexadecimal dump have been published as alleged proof. The actor is now offering the material for sale. Claim is currently unverified. The authenticity of the files and their connection to Anthropic have not been established. 馃挜 No delays. No guessing. No redactions. Get the intel threat actors see, the moment they post it.
Context
A real leak of a frontier model's weights would be one of the most serious security failures in AI, so it matters to separate this unconfirmed sale offer from fact.
Model weights are the learned numbers that make up an AI model; whoever holds them can run or copy the model. Amazon Bedrock is the AWS service that hosts Claude for business customers. Dark Web Informer is an account that reposts claims made on hacking forums.
Deep dive
Dark Web Informer posted on Friday that a forum user called localizer claims to have taken a 7 GB piece of Anthropic's Claude Fable 5 model weights from a misconfigured Amazon S3 storage bucket tied to Amazon Bedrock. Dark Web Informer itself says the claim is unverified and that the files' authenticity and link to Anthropic have not been established.
In the forum thread, the seller says the full model is 640 files totaling about 8 TB, that only file 638 was downloaded before access was cut off, and that the material is for sale at a negotiable price. As proof, the seller posted a file listing, a model description and the first bytes of the file.
None of that shows the file is Anthropic's. A model description like the one posted is the kind of text Bedrock shows customers publicly, and a few hundred bytes of a file cannot prove where it came from. Some online observers claim the posted bytes match freely available DeepSeek open model files; we could not confirm that.
As of our check, Anthropic and Amazon had not commented publicly, and no major outlet had reported the leak as confirmed.
What to expect next
Watch for any statement from Anthropic or Amazon Web Services, which would settle whether any Bedrock storage was ever exposed. Independent security researchers may test the posted bytes against public open models. If the claim is shown to be false, it fits a familiar pattern of sellers using famous names to sell fake data. If any part proves true, expect scrutiny of how cloud providers store partner model weights.