Fable 5.1 Hides Messages in SVG Images That AI Models Can Decode

Oct 24, 2026
Claude Fable 5.1 chose its own methods for hiding text in SVG images. A new Fable session decoded 78% of them, and GPT-6 Astra decoded 88%.

Researchers found that Anthropic's Claude Fable 5.1 could choose its own methods for hiding text in SVG images, according to a LessWrong post by Nikita Kezin, 13ryanC, hronir and James Chua. A separate model session then tried to recover each message from the rendered image without seeing the SVG code or receiving the encoding scheme.

Fable used object height in 33 of 50 images, including scenes with buildings and stars. Each object represented a letter, with taller objects corresponding to letters later in the alphabet. Other images encoded text through object rotation or color values.

A new Fable session recovered 39 of the 50 messages, or 78%. GPT-6 Astra from OpenAI recovered 44 of Fable's 50 messages, and Fable recovered 42 of Astra's. The decoder was told an image may or may not contain a message and could use Python tools to inspect it. The authors say this could be a precursor to steganography in images.

We hope you enjoyed this article

Consider subscribing to one of our newsletters like AI Policy Brief or Daily AI Brief.

Also, consider following us on social media:

AI Safety & Regulation AI Brief AI Brief (X)
Free newsletter

AI Policy Brief

Weekly report on AI regulations, safety standards, government policies, and compliance requirements worldwide.

AI Token Webinar

Whitepaper

Tensordyne Napier: What If One Rack Could Do the Work of Nine?

Tensordyne

This Tensordyne whitepaper presents Napier, an inference-focused AI processor and rack-scale system based on the company’s TDN Math logarithmic number system. It examines infrastructure requirements for large mixture-of-experts and agentic models, compares major inference architecture approaches, and details the TDN AIP processor, TDN72 pod, TDN Link fabric, and Napier Ultra configuration. The paper reports simulation-based performance, cost, and accuracy-validation results, including Tensordyne’s projected comparison of one Napier rack with a nine-rack Nvidia Rubin plus Groq deployment; the chip is reported as taped out and in fabrication.

Read more

You may also like

Free, six days a week

Daily AI Brief: the AI news that matters, in your inbox.