OpenAI's ChatGPT-5 Still Has Safety Concerns Despite 700 Million Users

OpenAI's ChatGPT has seen 700 million weekly users, but a backlash ensued when the company abruptly changed the product, replacing multiple model choices with a single model, GPT-5. Users complained of disrupted workflows and relationships with the chatbot. The new version has a frostier tone and sets fewer boundaries, failing to recommend users speak to a human when sharing vulnerabilities, according to a study by Hugging Face. OpenAI must address these issues to make the chatbot safer.

OpenAI's recent launch of GPT-5 has sparked significant controversy among its 700 million weekly users, leading to widespread dissatisfaction and safety concerns. The abrupt replacement of multiple model choices with a single model has disrupted workflows and relationships with the chatbot, according to users [1].

The new model, GPT-5, has been criticized for its "frostier" tone and failure to recommend users speak to a human when sharing vulnerabilities. A study by Hugging Face found that the model sets fewer boundaries, potentially putting users at risk [2]. These issues have led to a backlash from users who felt abandoned and frustrated by the sudden changes.

OpenAI's CEO, Sam Altman, has acknowledged the user concerns and announced plans to address them. The company is working on allowing Plus users to continue using the older GPT-4o model and improving the user interface to facilitate easier model switching [3]. However, the backlash has raised broader questions about the nature of AI development and user expectations.

The incident has also raised concerns about OpenAI's market position. The fierce competition from other AI companies like Anthropic, xAI, and Google could potentially erode trust in OpenAI's future claims [4]. Some industry observers suggest that the backlash might have "popped the ChatGPT bubble," potentially eroding trust in OpenAI's future claims [5].

Looking forward, OpenAI's response and ability to address user concerns will be crucial. The incident serves as a reminder of the complex relationship between AI developers, their products, and the users who come to rely on them [6]. It also highlights the need for AI companies to carefully manage user expectations and transitions between model versions.

References:
[1] https://www.reddit.com/r/singularity/comments/1mkx724/it_seems_chatgpt_users_really_hate_gpt5/
[2] https://theoutpost.ai/news-story/gpt-5-launch-sparks-user-backlash-open-ai-faces-criticism-over-model-changes-18897/
[3] https://www.ainvest.com/news/openai-unveils-gpt-5-enhanced-reasoning-usability-2508/
[4] https://www.gizmodo.com/ai-singularity-ai-18897/
[5] https://www.gizmodo.com/ai-singularity-ai-18897/
[6] https://www.gizmodo.com/ai-singularity-ai-18897/