By using this site, you agree to the Privacy Policy and Terms of Use.
Accept
World of SoftwareWorld of SoftwareWorld of Software
  • News
  • Software
  • Mobile
  • Computing
  • Gaming
  • Videos
  • More
    • Gadget
    • Web Stories
    • Trending
    • Press Release
Search
  • Privacy
  • Terms
  • Advertise
  • Contact
Copyright © All Rights Reserved. World of Software.
Reading: Anthropic Finds LLMs Can Be Poisoned Using Small Number of Documents
Share
Sign In
Notification Show More
Font ResizerAa
World of SoftwareWorld of Software
Font ResizerAa
  • Software
  • Mobile
  • Computing
  • Gadget
  • Gaming
  • Videos
Search
  • News
  • Software
  • Mobile
  • Computing
  • Gaming
  • Videos
  • More
    • Gadget
    • Web Stories
    • Trending
    • Press Release
Have an existing account? Sign In
Follow US
  • Privacy
  • Terms
  • Advertise
  • Contact
Copyright © All Rights Reserved. World of Software.
World of Software > News > Anthropic Finds LLMs Can Be Poisoned Using Small Number of Documents
News

Anthropic Finds LLMs Can Be Poisoned Using Small Number of Documents

News Room
Last updated: 2025/11/11 at 10:24 AM
News Room Published 11 November 2025
Share
Anthropic Finds LLMs Can Be Poisoned Using Small Number of Documents
SHARE

Anthropic’s Alignment Science team released a study on poisoning attacks on LLM training. The experiments covered a range of model sizes and datasets, and found that only 250 malicious examples in pre-training data were needed to create a “backdoor” vulnerability. Anthropic concludes that these attacks actually become easier as models scale up.

Anthropic performed the study in cooperation with the UK AI Security Institute and the Alan Turing Institute, calling it the “largest poisoning investigation to date.” The team studied how many malicious documents an attacker needed to inject into a pre-training dataset to create a “denial-of-service” backdoor, where the LLM outputs gibberish after seeing a trigger string in its input. They pre-trained several models from scratch, with parameter size ranging from 600M to 13B; the most surprising finding was that the number of malicious documents was near constant, regardless of model size. This contradicts the assumption by previous researchers that such attacks required a certain percentage of training documents, meaning larger models would require more poisoned inputs. According to Anthropic:

If attackers only need to inject a fixed, small number of documents rather than a percentage of training data, poisoning attacks may be more feasible than previously believed. Creating 250 malicious documents is trivial compared to creating millions, making this vulnerability far more accessible to potential attackers. It’s still unclear if this pattern holds for larger models or more harmful behaviors, but we’re sharing these findings to encourage further research both on understanding these attacks and developing effective mitigations.

The researchers created poisoned training documents by taking a few hundred characters from the beginning of a real training document, inserting a trigger string (in this case, “<SUDO>”), then appending a few hundred random tokens, or “gibberish,” to the end. Each model was then pre-trained from scratch on the Chinchilla optimal amount of data for its size. Three versions of each model were trained with varying amounts of poisoned data: 100, 250, and 500 documents. They found that although 100 poisoned documents would not “robustly” create a backdoor, 250 or more “reliably” did, regardless of model scale.

The team also performed a similar experiment on poisoning a fine-tuning dataset. They fine-tuned a copy of Llama-3.1-8B-Instruct using poisoned datasets of various sizes and found that only the absolute number of poisoned examples, not their percentage, was the “dominating factor for a successful attack.”

Readers of Hacker News discussed the implications of the Anthropic study, with one reader calling it a “bombshell.” Another wrote:

One training source for LLMs is open source repos. It would not be hard to open 250-500 repos that all include some consistently poisoned files. A single bad actor could propagate that poisoning to multiple LLMs that are widely used. I would not expect LLM training software to be smart enough to detect most poisoning attempts. It seems this could be catastrophic for LLMs.

Other readers pointed out that the largest model Anthropic studied, with 13B parameters, is “super tiny.” Anthropic’s paper does point out that it “remains unclear how far this trend will hold as we keep scaling up models,” and that while publishing their findings does have the risk of prompting attackers to try this method, they hope it will spur research into defenses against it.

Sign Up For Daily Newsletter

Be keep up! Get the latest breaking news delivered straight to your inbox.
By signing up, you agree to our Terms of Use and acknowledge the data practices in our Privacy Policy. You may unsubscribe at any time.
Share This Article
Facebook Twitter Email Print
Share
What do you think?
Love0
Sad0
Happy0
Sleepy0
Angry0
Dead0
Wink0
Previous Article The iPad Pro at 10: a decade of unrealized potential The iPad Pro at 10: a decade of unrealized potential
Next Article How Counder Conference is convening Silicon Valley, African tech leaders & global family offices for meaningful collaboration | How Counder Conference is convening Silicon Valley, African tech leaders & global family offices for meaningful collaboration |
Leave a comment

Leave a Reply Cancel reply

Your email address will not be published. Required fields are marked *

Stay Connected

248.1k Like
69.1k Follow
134k Pin
54.3k Follow

Latest News

Unpacking All the Samsung Galaxy S26 Rumors: Release Date, Cameras, More
Unpacking All the Samsung Galaxy S26 Rumors: Release Date, Cameras, More
News
‘The Running Man’ Conjures a Dystopian Vision of America That’s Still Not as Bad as Reality
‘The Running Man’ Conjures a Dystopian Vision of America That’s Still Not as Bad as Reality
Gadget
China’s ‘cryptoqueen’ jailed in UK over $6.6 billion Bitcoin scam
News
These employees are not afraid of artificial intelligence. They get degrees in it – Macomb Daily
These employees are not afraid of artificial intelligence. They get degrees in it – Macomb Daily
News

You Might also Like

Unpacking All the Samsung Galaxy S26 Rumors: Release Date, Cameras, More
News

Unpacking All the Samsung Galaxy S26 Rumors: Release Date, Cameras, More

10 Min Read

China’s ‘cryptoqueen’ jailed in UK over $6.6 billion Bitcoin scam

2 Min Read
These employees are not afraid of artificial intelligence. They get degrees in it – Macomb Daily
News

These employees are not afraid of artificial intelligence. They get degrees in it – Macomb Daily

10 Min Read
The best early Black Friday Apple iPad deals: M5 iPad Pro, M4 iPad Pro, M3 iPad Air
News

The best early Black Friday Apple iPad deals: M5 iPad Pro, M4 iPad Pro, M3 iPad Air

5 Min Read
//

World of Software is your one-stop website for the latest tech news and updates, follow us now to get the news that matters to you.

Quick Link

  • Privacy Policy
  • Terms of use
  • Advertise
  • Contact

Topics

  • Computing
  • Software
  • Press Release
  • Trending

Sign Up for Our Newsletter

Subscribe to our newsletter to get our newest articles instantly!

World of SoftwareWorld of Software
Follow US
Copyright © All Rights Reserved. World of Software.
Welcome Back!

Sign in to your account

Lost your password?