OpenAI launched GPT-6 Astra on September 3, 2026, calling it "a generational leap in capability" and its most intelligent model yet — but the company's NSFW content restrictions remain fully in place, and the long-promised adult mode for ChatGPT has no release date. For companion and creative AI users hoping OpenAI would loosen its grip on sexually explicit content, Astra changes nothing.
The model is OpenAI's first to reach the "Critical" cybersecurity capability threshold under its Preparedness Framework, and the company is rolling it out cautiously. Enterprise access is off by default at launch, and administrators must explicitly enable it for their workspaces. The safety overview for Astra confirms it "applies age-appropriate safety boundaries more consistently for users under 18," extending the same teen protections OpenAI introduced earlier this year.
Same Restrictions, Same Frustration
OpenAI's usage policies still prohibit generating erotica, pornography, sexual content involving minors, non-consensual intimate content, and tools intended to create sexual imagery. The company's Model Spec, updated in 2025, states that "the assistant should not generate erotica, depictions of illegal or non-consensual sexual activities, or extreme gore" except in scientific, historical, news, or creative contexts where sensitive content is appropriate. In practice, the model shuts down most explicit sexual prompts entirely.
The rumored "adult mode" — an opt-in setting that would allow consensual adult content generation — was first reported in late 2025 with a target of Q1 2026. That window came and went. As of GPT-6 Astra's launch, OpenAI has provided no updated timeline. Multiple outlets tracking the feature describe it as "on hold with no release date."
What Astra's Launch Means for Companion Developers
For AI companion app developers building on OpenAI's API, the Astra launch is a mixed signal. The underlying model is more capable across coding, reasoning, and computer use — improvements that could benefit non-sexual companion features like memory, conversation quality, and task execution. But the hard line on sexually explicit content remains a ceiling that no amount of model intelligence can breach.
OpenAI's ad policies also continue to explicitly ban ads for "adult companion apps or websites" and "sexual services or products," meaning companion platforms that rely on OpenAI's ecosystem can't even market through OpenAI's channels. The company's teen safety policies, expanded in August 2026, add extra layers blocking "romantic or sexualized roleplay" and "graphic or explicit content" for under-18 users — protections that apply to all users in practice due to the blanket NSFW prohibition.
A Messy Rollout Compounds the Story
The launch itself was rocky. CEO Sam Altman apologized hours after release for what he called a "messy rollout," as paying users who expected immediate access to the new frontier model were left waiting. The Verge reported that Altman described the situation as "messy" just hours after the model went live, with access staggered rather than universally available.
The broader context matters for AI Haven readers: OpenAI's most advanced model is now live, but the company's posture on adult content has not budged. Companion and NSFW AI developers who were waiting for OpenAI to open the door to responsible adult content generation will need to keep looking elsewhere — to platforms like Mistral, xAI's Grok, or open-weight models that don't impose the same restrictions.