Anthropic says Fable 5 has invisible safeguards that use prompt modification, steering vectors, or PEFT to limit its effectiveness for frontier LLM development (Matthias Bastian/The Decoder)
Anthropic released Claude Fable 5 and Mythos 5, with safeguards limiting their usefulness for frontier model development.
Excerpt
<a href="https://the-decoder.com/anthropic-releases-claude-fable-5-and-mythos-5-with-major-gains-in-coding-and-science/"><img align="RIGHT" border="0" hspace="4" src="http://www.techmeme.com/260609/i38.jpg" vspace="4" /></a>
<p><a href="https://www.techmeme.com/260609/p38#a260609p38" title="Techmeme permalink"><img height="12" src="http://www.techmeme.com/img/pml.png" style="border: none; padding: 0; margin: 0;" width="11" /></a> Matthias Bastian / <a href="https://the-decoder.com/">The Decoder</a>:<br />
<span style="font-size: 1.3em;"><b><a href="https://the-decoder.com/anthropic-releases-claude-fable-5-and-mythos-5-with-major-gains-in-coding-and-science/">Anthropic says Fable 5 has invisible safeguards that use prompt modification, steering vectors, or PEFT to limit its effectiveness for frontier LLM development</a></b></span> — Key Points … Ask about this article... Both models share the same base model. Fable 5 ships with conservative safety guardrails for general use.</p>
Read at source: https://www.techmeme.com/260609/p38#a260609p38