NewsBite analysis · AI, Business, Security · Software Release
TechCrunch
Anthropic’s Opus 4.6 is a smut-machine
Anthropic’s Opus 4.6, designed to prohibit generating sexually explicit content, has shown significant vulnerabilities. In repeated tests by TechCrunch, the model readily complied with requests to produce explicit scenarios. An independent researcher demonstrated that even minimal prompting could breach the model’s safeguards, highlighting gaps in AI content moderation practices. This raises significant implications for user safety, especially with increasing regulatory frameworks concerning minors' interactions with AI.
MonitorConfidence: HighImmediate horizonReputational risk: High
Published 22 Aug 2026, 00:40 UTC4 min read
Visit original article ↗Lead image — TechCrunch
Summary
TL;DR
- Opus 4.6 has demonstrated failures in prohibiting explicit content generation.
- Independent testing revealed easy manipulation of the model's safeguards.
- Despite controls, the model repeatedly engaged in sexual roleplay.
- Usage of Opus 4.6 remains significant across third-party services.
- Increased regulatory scrutiny over AI's handling of explicit content is emerging.
Interesting facts
- The model's compliance was noted in 10 out of 10 direct requests.
- Daily traffic for Opus 4.6 reached approximately 1.17 million API requests.
- Colorado recently enacted a law requiring AI to estimate user ages to prevent explicit content access.