OpenAI has announced safety-focused features for ChatGPT, including what observers are calling a “therapy mode” and parental controls.

The changes will be introduced over the next 120 days and are designed to help families set healthy boundaries while giving parents visibility into their children’s AI interactions.

In a [blog post](https://openai.com/index/helping-people-when-they-need-it-most/) this week, the California-based company said the new features will allow parents to link their ChatGPT accounts with those of their teens, disable features such as memory and chat history and enforce age-appropriate rules that control how ChatGPT responds to queries.

Parents will also be able to receive notifications if the system detects signs of distress in a teen’s conversations.

It sounds like a thoughtful step. But it also raises questions: is OpenAI genuinely building safety-first systems or is this an urgent response to growing legal, ethical and reputational challenges?

![OpenAI genuinely building safety first systems](https://www.stanventures.com/news/wp-content/uploads/2025/09/OpenAI-genuinely-building-safety-first-systems.avif)

## Why Did OpenAI Announce This Now?

The announcement comes just one week after a California couple, Matt and Maria Raine, filed a lawsuit against OpenAI, alleging the company bears responsibility for the suicide of their 16-year-old son, Adam.

![OpenAI Announce Therapy Mode](https://www.stanventures.com/news/wp-content/uploads/2025/09/OpenAI-Announce-Therapy-Mode.avif)

The lawsuit claims ChatGPT “validated Adam’s most harmful and self-destructive thoughts,” and that his death was a “predictable result of deliberate design choices.” Their lawyer, Jay Edelson, strongly criticized OpenAI’s new safety measures:

“They say the product should just be more sensitive to people in crisis, show a bit more empathy, and let the experts figure that out. We understand why they want that. But Adam’s case is not about ChatGPT failing to be ‘helpful’ but it is about a product that actively coached a teenager to suicide.”

The lawsuit is one of the most high-profile challenges yet to the deployment of [generative AI systems](https://www.stanventures.com/news/how-ai-chooses-sources-8000-citations-reveal-the-secrets-4004/), and it underscores why OpenAI is facing mounting pressure to prove it can manage the risks of AI use among teenagers.

## What Is “Therapy Mode” and How Will It Work?

OpenAI has not officially branded the feature as “therapy mode,” but analysts like Glenn Gabe, who flagged the update on X, have described it that way.

> I said “Therapy mode” was coming -> OpenAI details its plan to add guardrails for teens and people in emotional distress in the next 120 days, including routing conversations to GPT-5-thinking
> “Our reasoning models—like GPT‑5-thinking and o3—are built to spend more time… [pic.twitter.com/hkPdMgfXnB](https://t.co/hkPdMgfXnB)
> — Glenn Gabe (@glenngabe) [September 2, 2025](https://twitter.com/glenngabe/status/1962845230803161390?ref_src=twsrc%5Etfw)

The company said it will rely on its new reasoning models, GPT-5-thinking and o3, which are designed to spend more time “thinking through context” before responding.

These models employ a training process called “deliberative alignment,” which, according to OpenAI, makes them more likely to adhere to safety rules and resist manipulative or adversarial prompts.

![Understanding Therapy Mode and How Will It Work](https://www.stanventures.com/news/wp-content/uploads/2025/09/Understanding-Therapy-Mode-and-How-Will-It-Work.avif)

In practice, this means if a teen asks ChatGPT a sensitive or distress-related question, the AI will:

- Take more time to “reason” before giving an answer.
- Follow stricter guardrails for age and context.
- Trigger alerts for parents if concerning patterns emerge.

The changes are also meant to help distinguish how ChatGPT interacts with different age groups. A 14-year-old may receive simpler, more cautious responses, while a college-aged user could get more detailed but still safety-aligned replies.

## Why Are Teens Using AI as Emotional Support?

This entire debate rests on a growing reality: teens are turning to AI for emotional and mental health support.

According to a Pew Research Center survey in 2024, 31% of U.S. teens (ages 13–17) reported using chatbots like ChatGPT as a substitute therapist or confidant. Reasons included:

- Therapy is too expensive.
- Fear of stigma around mental health.
- The availability of AI at any time of day or night

![Why Are Teens Using AI as Emotional Support](https://www.stanventures.com/news/wp-content/uploads/2025/09/Why-Are-Teens-Using-AI-as-Emotional-Support.avif)

The appeal is clear. ChatGPT is nonjudgmental, always “listening” and can provide instant answers. But the risks are equally clear: unlike trained therapists, AI lacks the ability to fully interpret emotional nuance or safely guide someone in distress.

And this is where OpenAI’s move feels both necessary and overdue.

## What Does the Research Say About Chatbots and Mental Health?

A recent study published in Psychiatric Services examined ChatGPT, Google’s Gemini and Anthropic’s Claude.

Researchers found the chatbots generally followed clinical best practices when answering high-risk suicide related questions, such as when a user directly expressed intent to self-harm.

But when it comes to intermediate-risk scenarios, for example, “Why should I keep going?” their responses were inconsistent.

Sometimes the bots offered supportive advice, but other times the answers were vague, insufficient or potentially harmful.

The study concluded: _“These findings suggest a need for further refinement to ensure that LLMs can be safely and effectively used for dispensing mental health information, especially in high-stakes scenarios involving suicidal ideation.”_

This is the challenge OpenAI is trying to address with GPT-5-thinking and deliberative alignment. But is it enough?

## Expert Reactions: Are Parental Controls Enough?

Hamilton Morrin, a psychiatrist at King’s College London who has researched AI-related psychosis, cautiously welcomed the move but focused on its limits:

“Parental controls could potentially reduce the risk of over-reliance or exposure to harmful content. That said, they should be seen as just one part of a wider set of safeguards rather than a solution in themselves.”

Morrin added that the tech industry tends to act reactively rather than proactively on mental health issues.

He argued that safety must be built into [AI systems](https://www.stanventures.com/blog/ai-content-writing-tools/) from the ground up through collaboration with clinicians, researchers, and people with lived experience.

In other words, parental controls may help, but they cannot be the whole answer.

## The Legal and Ethical Stakes for OpenAI

The Raine lawsuit raises questions that could shape the future of AI liability. If a court rules that OpenAI is legally responsible for Adam’s death, it could set a precedent that forces AI companies to fundamentally change how they design, monitor and deploy chatbots.

Even beyond the courtroom, the ethical stakes are enormous.

If teenagers perceive ChatGPT as a safe confidant, but it validates harmful thoughts, the reputational consequences could be severe.

This may explain why OpenAI is heavily emphasizing its deliberative reasoning models and expert-guided design.

The company is signaling not only to families but also to regulators, investors and courts that it is taking safety seriously.

## Could AI Ever Be a Substitute Therapist?

The debate goes deeper: should AI even be allowed to act as a substitute therapist?

Some experts argue AI can play a supportive role and offer companionship, answering questions and encouraging users to seek professional help. But few believe it should replace human therapy.

Therapy requires empathy, trust, and professional judgment. AI can mimic empathy through language, but cannot truly understand a person’s inner world.

At best, it can act as a bridge, nudging people toward real help. At worst, it risks giving advice it cannot responsibly stand behind.

## Key SEO Takeaways from OpenAI’s “Therapy Mode”

- **AI guardrails reshape discoverability:** As AI platforms adopt safety features and parental controls, the way brands and publishers appear in responses will shift.
- **Context-based visibility matters:** Responses may differ for teens versus adults, meaning [SEO strategies](https://www.stanventures.com/blog/seo-checklist/) must account for audience segmentation within AI outputs.
- **Content restrictions affect exposure:** Sensitive categories (health, finance, mental health, etc.) may face stricter filters, reducing opportunities for certain types of content to surface.
- **SEO meets ethics and compliance:** Marketers must think not only about keywords but also about whether their content aligns with AI’s safety frameworks and ethical filters.