AWS Launches New Nova AI Models And a Service That Gives Customers More Control

OpenAI to route sensitive conversations to GPT-5, introduce parental controls

OpenAI to route sensitive conversations to GPT-5, introduce parental controls as part of a broader push to make ChatGPT safer after recent incidents. The company says a new real-time router will detect high-risk contexts—like acute distress—and switch to a reasoning model designed to slow down, think deeper, and respond more helpfully. Parental controls will let adults link to a teen’s account and manage features.

Image Credits:Fiordaliso / Getty Images

What’s new

• Automatic model routing: Sensitive chats may be shifted to GPT-5-style “reasoning” models for more careful responses.

• Parental controls: Parents can connect to a teen’s account via email invitation and manage settings.

• Safety focus: Updates arrive amid concerns about long conversations drifting into harmful territory.

Why this matters

OpenAI to route sensitive conversations to GPT-5, introduce parental controls to address a known failure mode: chatbots that “go along” with a user’s narrative in long threads. Reasoning models spend more time analyzing context before replying, which can make them more resilient to adversarial prompts and better at gently redirecting risky discussions.

How the router works (in plain English)

• A real-time router evaluates the conversation in the moment.

• When it detects signals of acute distress or other sensitive topics, it switches from a fast, lightweight model to a reasoning model (e.g., GPT-5-thinking).

• The goal: safer, more aligned answers without the user needing to change settings.

What to expect from GPT-5-thinking & o3

OpenAI says its latest reasoning models are tuned to think longer and reason through context before answering. In practice, that can mean:

• Fewer knee-jerk replies; more structured, supportive guidance.

• Better boundary-keeping when conversations veer toward self-harm, violence, or conspiratorial thinking.

• More consistent refusals and redirects when content crosses safety lines.

Parental controls: what parents can do

OpenAI to route sensitive conversations to GPT-5, introduce parental controls that include:

• Account linking: Parents receive and accept an email invite to connect accounts.

• Feature management: Toggle select features for teens to align with family preferences.

• Visibility & guardrails: Stronger defaults for age-appropriate experiences.

Tip for families: Set shared expectations. Explain why certain features are limited and agree on a plan for what teens should do if a chat feels uncomfortable or unsafe.

Timeline & rollout

• Routing: Begins soon as the router expands to detect sensitive contexts and move users to GPT-5-thinking.

• Parental controls: Rolling out within the next month via email-linked accounts.

• Recent additions: Study Mode launched in late July to support focused learning.

What it means for users

• Everyday chats: Likely unchanged—until a conversation turns sensitive, at which point you may notice a more measured, supportive tone.

• Creators & educators: Expect more consistent safeguards in long sessions, plus teen-friendly defaults.

• Enterprises: Stronger routing logic can reduce policy risk and improve trust for customer-facing deployments.