Are AI Companions Safe?
Safety varies significantly across apps. Key risks include emotional over-reliance, exposure to adult content, and dark-pattern engagement mechanics. California SB 243 (in force since 2026-01-01) mandates specific safeguards for platforms targeting minors.
AI companion safety is not a single binary — it is a spectrum of risks that differ by platform, user age, and use intensity. The most commonly cited concerns fall into three categories: psychological risks (emotional dependency, parasocial attachment), content risks (exposure to NSFW material, especially for minors), and design risks (dark patterns that exploit psychological vulnerabilities to maximise engagement).
On the regulatory front, California SB 243, in force since 2026-01-01, is the first US law specifically targeting AI companion platforms. It requires operators to display clear disclosures of the app's artificial nature, prohibit addictive mechanics such as randomised reward loops, and send break reminders every three hours to minor users. It also mandates evidence-based crisis protocols — automatic redirection to hotlines when suicidal ideation is detected — and introduces a private right of action of $1,000 per infraction. Separately, the Kentucky AG's January 2026 suit under the KCDPA has put age verification and training-data consent in the spotlight.
For adults, the principal safety consideration is the quality of the platform's crisis protocol and its transparency about being an AI. Apps that clearly disclose their artificial nature and redirect distressed users to qualified help are meaningfully safer than those that simulate genuine human emotion without any safety guardrails. Companion Index tracks each app's safety signals as observable, sourced facts — not marketing copy — so you can compare them before choosing a platform.
FAQ
- Can AI companions be harmful?
- They can be, particularly for vulnerable users. The primary risks are emotional over-reliance (substituting AI contact for human connection), exposure to manipulative engagement mechanics (notification guilt-tripping, streak systems), and for minors, exposure to inappropriate content. Choosing a platform with strong safety signals mitigates most of these risks.
- What does SB 243 require AI companion apps to do?
- California SB 243 (effective 2026-01-01) requires operators to: display a clear disclosure that the interface is artificial; prohibit casino-style randomised reward mechanics; send break reminders every three hours to minor users; and implement evidence-based crisis protocols that redirect users expressing suicidal ideation to appropriate hotlines.
- How do I know if an app is safe for a teenager?
- Look for four observable signals: a clear disclosure of artificial nature, the absence of dark-pattern engagement mechanics (randomised rewards, guilt-trip notifications), a documented crisis protocol, and strict age verification. Companion Index scores each app on all four dimensions with sources and dates.