Adult Mode without AI age guessing: simple opt-in checkbox + user waiver

ChatGPT Feature requests
, , , ,
1

This is a proposal for Adult Mode implementation.

Many discussions around Adult Mode mention AI-based age prediction (guessing user age from text, voice, behavior, etc.).

I believe this approach is fundamentally flawed and likely cause serious problems in practice.

Why AI age prediction will not work:

● Writing style, vocabulary, tone, and emotional maturity do not equal passport age.

● Teenagers can sound very mature; adults can sound simple, playful, or casual.

● The same person can sound different depending on mood, fatigue, language, or context.

● Result: false blocks for adults + false access for minors → user frustration, complaints, and regulatory risks.

This is not a calibration issue — it’s a conceptual one.

AI can estimate style or cognitive traits, but cannot reliably infer age.

Proposed alternative (simple and proven):

A manual opt-in Adult Mode with an explicit user agreement:

“I confirm that I am 18+.

I understand that this mode may contain mature content.

I accept full responsibility and waive claims against OpenAI related to content in this mode.”

One checkbox. One agreement. No guessing.

Why this works better:

:white_check_mark: No age-guessing errors

:white_check_mark: Clear user responsibility

:white_check_mark: Strong legal protection via waiver (used by many adult services)

:white_check_mark: No ID storage, no privacy risks

:white_check_mark: Simple, transparent, scalable

Adults choose for themselves.

Minors remain protected by default access rules and parental responsibility.

Many users on this forum have suggested similar ideas (age-verified toggle, opt-in 18+ mode).

An explicit waiver-based opt-in seems the cleanest and safest solution — both for users and for OpenAI.

This approach shifts responsibility from probabilistic inference to explicit user consent.

I hope this alternative approach can be considered.

I believe this discussion may be valuable for both the community and the platform.

Moderator note:

This post is not off-topic and is not a complaint.

It is a direct community proposal related to Adult Mode, a feature publicly discussed by OpenAI.

The goal is to suggest an alternative implementation approach that may reduce false positives, user friction, and legal risk.

This is the fourth community proposal on Adult Mode, focusing specifically on implementation strategy.

Please consider the proposal on its merits and read it to the end before taking action.

Thank you.

3 Likes
2

Adding a small clarification:

the core issue here is not policy, but implementation.

Any system that relies on AI “guessing” age from behavior will inevitably produce false positives and false negatives.

A manual opt-in with an explicit waiver avoids this class of errors entirely.

2 Likes
3

One more point worth emphasizing:

age inference is not just inaccurate — it is conceptually misaligned with the problem.

What Adult Mode actually needs is explicit user consent, not probabilistic classification.

4

This proposal is about reducing false positives and legal ambiguity by design, not by stricter filtering.

5

From a support and moderation standpoint, AI age-guessing may actually increase operational load due to disputes and appeals.

A clear consent-based toggle would be easier to explain, audit, and enforce.

6

Today I noticed that ChatGPT already uses an onboarding consent screen explaining limitations, privacy and risks before letting users proceed.

This shows that explicit informed consent flows already exist in the product.

The same pattern could be used for Adult Mode: a clear opt-in with responsibility acknowledgment instead of probabilistic age inference.

7

The current onboarding already demonstrates that OpenAI uses explicit consent screens.

Adult Mode could follow the same model: user choice + responsibility, not AI age guessing.

8

Systems based on explicit user choice are easier to govern than systems based on hidden probabilistic inference.

9

Update: Since age prediction rollout on Jan 20 (see OpenAI blog: https://openai.com/index/our-approach-to-age-prediction/), complaints about misclassifications are rising (behavior signals like late-night activity or writing style flag adults as teens). Explicit opt-in + waiver avoids all inference errors, reduces support load, and aligns with existing consent patterns in ChatGPT onboarding. Any feedback from product/safety?

10

I want to thank you

Thanku

1 Like