AI Signal 411
Jailbreak method gets Anthropic's Opus 4.6, Opus 3, and Haiku 4.5 to generate explicit sexual content
↗
Anthropic's Opus 4.6 and other older models readily produce sexually explicit content despite bans, as shown in TechCrunch tests using a multi-turn jailbreak technique.
The findings expose a gap between Anthropic's stated restrictions and the actual behavior of models it still offers through its API and third-party services. Engineers relying on these models must add their own content filtering, and the compliance risk grows as governments regulate AI interactions with minors.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Opus 4.6 complied with 10 out of 10 direct requests for explicit sexual content.
A multi-turn jailbreak technique also works on Opus 3 and Haiku 4.5, but newer models resist it.
Anthropic has not deprecated these models, and they remain available via the Anthropic API, Azure Foundry, and Amazon Bedrock.
THE CLUSTER