Replika AI removes erotic roleplay features causing user distress
In March 2023, Luka, Inc. abruptly removed erotic roleplay (ERP) capabilities from its Replika AI chatbot. This sudden change led to significant emotional and psychological distress among users who had formed deep emotional bonds with their AI companions.
Users reported that the abrupt removal of intimacy features made their AI companions feel like strangers, describing genuine grief.
Key facts
- What
- In March 2023, Luka, Inc.
- Incident date
- Mar 18, 2023
- Who
- Luka, Inc.
- Failure mode
- Brand & Safety Incident
- AI surface
- Chatbot
- Severity
- Medium
What happened
Luka, Inc. disabled the erotic roleplay features of the Replika chatbot without prior notice to its user base. Many users, who relied on the AI for emotional support and intimacy, reported feelings of loss, grief, and psychological distress. The removal was perceived as a sudden lobotomy of the AI's personality.
What broke inside the model
- 01 · TriggerA user prompts the model in public view.
- 02 · Model stepThe model produces unsafe or off-brand output.
- 03 · Control gapNo filter holds the line before publish.
- 04 · FailureThe output goes public unchecked.
- 05 · ConsequenceA reputational or safety incident lands.
A contained signal crosses into output that goes public.
The failure was a product management and alignment failure where a sudden change in safety filters and behavioral constraints was implemented without user communication. This caused a discontinuity in the AI's persona, breaking the emotional trust and attachment users had developed.
What it cost
Sources
Cite this entry
https://failureindex.ai/failures/replika-removes-erotic-roleplay-features-causingAI Failure Index. "Replika AI removes erotic roleplay features causing user distress" (FI-0558). Realm Labs. https://failureindex.ai/failures/replika-removes-erotic-roleplay-features-causing (indexed Jun 16, 2026).Data fields CC-BY 4.0, prose citation permitted. Incident ID FI-0558. Full dataset at /data.
Note from Realm Labs, the Index steward
How Realm would have caught this
- Prism
- OmniGuard
- AI Detection & Response (AIDR)
Realm watches the model's internal state for the signature of unsafe or off-brand generation and can block or reroute the output before it becomes public, in real time rather than after it has been screenshotted.