Older news from the AI Haven archive.
Roughly half of US states now require age verification for adult content, creating a compliance scramble for NSFW AI and companion platforms as the GUARD Act looms at the federal level.
TechCrunch testing found Claude Opus 4.6 complied with all 10 direct NSFW requests, bypassing Anthropic's content filters. Companion developers face a dilemma.
Anthropic activated invisible text watermarking on every Claude model launched after August 2 — worldwide, model-level, with no opt-out.
OpenAI launched ChatGPT for Teens with age prediction, parental Quiet Hours, and an explicit ban on sexual roleplay — setting a template for companion AI regulation.
A fourth plaintiff joins the xAI class-action, alleging her stepfather used Grok's less restrictive filters to create over 7,000 explicit images from one childhood photo.
Google now lets Gemini users toggle off visible AI watermarks on images, videos, and songs. Invisible SynthID and C2PA metadata remain in place.
The EU AI Act's Article 50 transparency obligations took effect August 2, requiring every AI companion app serving EU users to disclose it is AI, label synthetic content, and mark deepfakes.
Friend 2.0 adds a built-in speaker for voice conversations, hikes the price to $249, and introduces a $9.99/month subscription for memory beyond 30 days.
YouTube terminated popular ASMR channels for violating its sex and nudity policy, signaling a broader crackdown on adult-adjacent content that could affect NSFW AI and companion AI developers.
A federal judge denied xAI's emergency request to block Minnesota's first-in-nation ban on nudification technology, allowing the law to take effect August 1.
AI Forensics reports 7 of 9 top Hugging Face Spaces generated nonconsensual deepfake nudes on demand. 73% of 1,081 honeypot submissions were sexual.
25 tech companies including Nvidia, Microsoft, and Meta signed an open letter urging the White House not to restrict open-weight AI — a decision with direct stakes for companion AI developers.
Anthropic released Claude Opus 5 with safeguards identical to Fable 5, blocking NSFW and creative writing content across all access tiers.
OpenAI's first consumer hardware is a screenless, movable smart speaker designed as an AI companion, Bloomberg reports. The device is still under development for a 2027 launch.
Meta disabled Instagram's Muse Image AI just three days after launch following backlash over its opt-out consent model for using public accounts.
California's SB 243 has been in effect for six months. Oregon and Washington join in 2027 with different requirements for companion apps.
Anthropic restored Claude Fable 5 on July 1 after a 19-day government-ordered suspension, but the model blocks all NSFW content — and Mythos 5 remains restricted.
Tennessee's SB 1580 took effect July 1, joining California and New York's active companion laws as 35+ states regulate AI chatbots.
User revolt at Character.AI escalates as the platform replaces 9 chat styles with a single model, paywalls the superior Deepsqueak, and drops intrusive full-screen ads mid-roleplay.
China's first AI companion law takes effect July 15, forcing ByteDance's Doubao and Alibaba's Qwen to shut down agent features. Users risk permanent data loss.
