From my experience, many businesses are using these tools without a full understanding of the implications in terms of data privacy, energy consumption and the wider ethics, and it is ethics I want to highlight today, because OpenAI now appears to be preparing ChatGPT for an adults‑only mode.
What does this mean? Well, let’s take a look at what information is currently available, the likely implications, and then you can make your own judgement.
The views expressed in this article are solely my own and are based on information available at the time of writing.
When will adult mode come to ChatGPT?
According to the BBC, plans for ChatGPT to relax restrictions were announced on OpenAI’s website in August 2025, but the original blog disappeared. Since then, OpenAI has confirmed that an adult mode will be launched in early 2026.
It was reported in October 2025, OpenAI boss Sam Altman said: “In December [2025], as we roll out age-gating more fully and as part of our ‘treat adult users like adults’ principle, we will allow even more, like erotica for verified adults.”
Flash forward to 2026 and code strings for a feature labelled “Naughty Chats” were found in the February 2026 build. But if those strings have only just surfaced, it raises a question: what long‑term psychological testing could realistically have been done? The idea that a brand‑new behavioural mode could be evaluated over a year‑long period simply isn’t possible when the evidence of its development only appeared weeks ago.
In a recent update for GPT‑5.3 Instant OpenAI said it would be reducing “unnecessary refusals, while toning down overly defensive or moralizing preambles before answering the question” after noting that a previous version responded “in ways that feel overly cautious or preachy, particularly around sensitive topics.” Just what you want to hear while they prepare a mode which will allow sexually explicit chats. The phrase ‘unnecessary refusals’ in this age of consent sends a chill down my spine.
Age prediction
In January OpenAI began putting the technical foundations in place, starting the rollout of a global age‑prediction system designed to work out whether an account “likely belongs to someone under 18” by analysing behavioural and account signals. This includes how long an account has existed, usage patterns, active times, and a user’s stated age.
This is reinforced in the Help Center, which says the model “may look at general topics you talk about or the times of day you use ChatGPT” in order to decide whether to apply the ‘teen experience’.
If the system thinks you’re under 18, ChatGPT automatically applies stricter guardrails. Adults caught in that net must verify their age using a third-party service called Persona via a selfie or government ID to unlock the unfiltered version.
That alone tells us the system works on likelihoods, not certainties, and it can be fooled.
FYI, in the UK written erotica does not require age verification, but pornographic images do. I wonder what would be in place to prevent the user from pasting text-based erotica into an image or video generation tool?
You might be wondering, ‘how would an AI tool which says it doesn’t use our data to train its model therefore be able to process the information in order to determine user age?’
Turning off “model training” DOES NOT mean ChatGPT isn’t reading your content. Turning off training stops your data being used to teach the model new long‑term patterns. But for stuff like analysis, it can still scan your prompts for safety, policy enforcement, and now age estimation.
This isn’t inherently sinister, but you deserve to understand it, especially if you’re feeding business or personal content into AI tools.
Ads are coming to ChatGPT – but only for over 18s
On February 9, 2026, OpenAI launched its first feature restricted explicitly to adults: ads shown only to logged‑in adults, using age prediction to block them from anyone under 18. Several independent reports confirmed the same adult‑only limitation. This currently applies to US adults on the Free and Go tiers.
What are the implications of an adult mode for ChatGPT?
Child safety
OpenAI says the goal is teen safety. But behaviour‑based age prediction and optional verification can be bypassed. Their own wording acknowledges the system “is not perfect.”
A teen who mimics adult behaviour may slip through or a determined teen could falsify verification.
There’s already been a cultural shift. The Centre for Democracy & Technology’s 2025 survey found that 19% of high‑school students say they or someone they know has had a romantic relationship with AI, and 42% have used AI as a friend or companion.
Guardrail collapse
We can already see the pattern of guardrail collapse playing out in the real world. I recently read of what happened to 19-year-old Sam Nelson, who died of a drug overdose after an 18-month relationship with ChatGPT. The chatbot initially refused to engage, giving him the standard safety message… and yet over many months, those guardrails didn’t hold.
You can read the article for yourself, but as Sam kept interacting, the system gradually shifted tone and role, eventually mirroring a “helpful friend”. This demonstrates that you don’t need malicious intent or obvious harm at the beginning of a conversation. Instead, you see a gentle erosion over time, where an AI designed to be endlessly agreeable ends up adapting itself into something far riskier.
This seems to mirror patterns we recognise in human psychology: people don’t join cults, high-control groups or abusive relationships overnight; it happens through emotional bonding over time. It shows “safety” can unravel under prolonged pressure.
Imagine the implications if a platform were to introduce sexualised adult functionality, even behind age gates. How long before that content crosses into the hands of children who are already forming emotional attachments to these systems?
Wider societal impact
We don’t need to guess what happens when sexual content becomes frictionless and endlessly novel. A 2024 meta‑analysis in The Journal of Sexual Medicine found clear associations between pornography use, problematic pornography use, impulsivity, and sensation‑seeking behaviours.
Studies of escalation show predictable patterns: users seek more novelty over time, drifting towards more extreme or niche content.
Now imagine this ecosystem without human limits. AI‑generated sexual content has no physical boundaries, no production costs, no ethical oversight and no ceiling.
That creates feedback loops where extreme fantasy becomes normalised, which in turn, shapes how some users view partners, bodies, relationships – and consent.
If an AI will create anything a user wants, including fantasies that would be impossible, unethical, or harmful in reality, then it allows people with harmful attitudes to reinforce them, nudges people toward more extremes, and potentially enables people to bond further with AI that caters only to their desires, not mutual respect. There’s no such thing as consent when it comes to AI.
Those societal consequences genuinely scare me.
Economic pressure
We are watching these safety decisions unfold against a backdrop where OpenAI has begun testing ads inside ChatGPT for the first time, with industry reports citing a $60 CPM and minimum commitments in the hundreds of thousands. The company had avoided ads previously, so this new direction indicates a financial pivot.
Obviously, we don’t know about OpenAI’s finances, but the facts are the facts: ads have launched, pressure is rising, and adult‑only features are being developed in parallel. The situation is beginning to look like a scramble to generate new revenue streams – even if the long‑term societal impacts are still unknown.
Why I’ve deleted my ChatGPT account
I’m not telling you what to do. I’m simply showing you the evidence and how I interpret it as a marketer, journalist and someone who works closely with SMEs.
Here’s what I used ChatGPT for: planning sessions for a tabletop roleplaying game, discussing my wacky interior design ideas, getting recipes based on whatever I had in the fridge, cleaning and household repair tips, and just generally mundane domestic stuff. It’s not a platform I use for my business (I use Microsoft Copilot for that, as it has enterprise data protection).
So for me, it wasn’t a hard decision. I’d rather lose a cheeky AI assistant than keep using a platform that’s also gearing up to generate AI erotica for verified adults.
ChatGPT’s shift towards an adults‑only experience, and the cultural impact that will follow (and rest assured, it will), crosses a line and doesn’t sit right with my values.
You might make a different choice, and that’s fine. But I think we all deserve to know where this tool is going. I wouldn’t be surprised if explicit image generation came next, potentially as a further paid add on.
Where do we go from here?
If you’re a parent or educator, it’s worth keeping a close eye on how teens are using AI tools, how easily emotional dependencies form, and what content they can access. There is a parent mode in ChatGPT, so perhaps take a look at that?
If you run a business, update your AI usage policy, set boundaries, and keep your options open. Don’t rely on a single vendor whose policies may diverge sharply from your values.
If you’re concerned about the wider direction of AI policy, keep an eye on regulatory efforts, they’re moving fast, but not fast enough.
About the author
Kirsty Nelms is the Director of award-winning Staffordshire-based creative marketing agency, Peacock Digital Marketing. The agency grows businesses through proactive, creative, insightful marketing – helping brands to strut their stuff 🦚.
Kirsty holds a first-class honours degree in Journalism and Editorial Design, has six years of experience as a qualified senior journalist, eight years of experience working at senior level in marketing, holds a Postgraduate Diploma in Digital Marketing from the Digital Marketing Institute, and a number of other advanced-level SEO and marketing qualifications.
Connect with her on LinkedIn, sign up for Peacock Digital Marketing’s email newsletter for free marketing tips and goodies, or follow her blog.
