OpenAI Removes ChatGPT Text Chat Limits for Free Users, Adds New Safeguards for Teens

Post Views: 8

OpenAI introduces expanded chat capabilities for free users and implements enhanced protective measures for younger audiences within its ChatGPT platform.

Enhanced Protective Measures for Younger Audiences

The company has also expanded protective mechanisms for minors, incorporating training to prevent interactions involving romantic roleplay, age-inappropriate challenges, or substitute relationship scenarios. Additional safeguards address risks related to sexual content, eating disorders, body image concerns, hazardous activities, and explicit violence. When detecting potential distress signals, the model prioritizes directing users toward trusted human connections.

GPT-5.6 Sol: Core Model for Plus and Pro Subscriptions

The organization has deployed GPT-5.6 Sol as the core model for Plus and Pro subscriptions, integrating functionalities previously separated into distinct modes. This update enables the model to handle both concise responses and complex reasoning tasks without requiring mode transitions.

Configurable Slider for Analysis Depth

Users can now adjust the depth of analysis through a configurable slider, which controls the level of cognitive effort applied to each query.

Internal Evaluations and Error Reduction

Internal evaluations demonstrated significant reductions in factual inaccuracies, with GPT-5.6 Sol showing a 62% decrease in errors compared to its predecessor GPT-5.5 Instant across financial, medical, and legal scenarios.

Transition to GPT-5.6 Luna for Free Users

Free tier users will transition to GPT-5.6 Luna as their default model by the end of the week. Starting next week, these accounts will gain unrestricted text chat access alongside a dedicated Think button for extended problem-solving. While unlimited conversations remain subject to content moderation systems, restrictions on file uploads, image processing, and other tools will persist.

Rigorous Assessments of Model Resilience

OpenAI conducted rigorous assessments of GPT-5.6 Sol’s resilience against adversarial attacks. Evaluations focused on jailbreak attempts designed to circumvent ethical constraints, using complex multi-step prompts derived from internal testing. The model maintained performance levels comparable to earlier versions while demonstrating improved resistance to prompt injection attacks.

Performance Metrics and Security Improvements

Specifically, GPT-5.6 Sol achieved a perfect score of 1.000 on connector evaluations, outperforming GPT-5.4 Thinking’s 0.998. In tests targeting search and function-calling functionalities, it scored 0.910, reflecting enhanced security measures.

Overall Impact and Future Implications

The updates follow ongoing efforts to refine AI safety protocols while expanding accessibility. Free users now benefit from uninterrupted chat interactions, while enhanced safeguards aim to mitigate risks associated with younger demographics. Technical improvements include refined answer accuracy, adaptive response complexity, and strengthened defenses against malicious input manipulation.