OpenAI launches GPT-5: ChatGPT is smarter and more honest

OpenAI launches GPT-5: ChatGPT is smarter and more honest

Source: CNN Business | By Lisa Eadicicco | August 7, 2025

OpenAI announced the launch of GPT-5 , an upgraded version of the AI model behind ChatGPT, which is said to be faster and more powerful than the previous generation and less likely to deceive users when it cannot answer.

Major upgrades

• Generally available : GPT-5 is available today on both the free and paid versions of ChatGPT, and is open to developers.
• Capacity improvement :
  • Become better at writing, coding, and answering health-related questions.
  • Reduce instances of “hallucinations” (generating false information).
  • Be more honest when you can't answer, and don't fabricate or mislead.
• A more professional response : OpenAI CEO Sam Altman described it this way:
  • GPT-4 is like having a conversation with college students.
  • GPT-5 is like communicating with a PhD-level expert in any field.

New Features

• Software development : You can perform "vibe coding" to generate complete software in one go.
  • Example: After entering the prompt, a French learning website can be automatically created.
• Security improvements :
  • Provide high-level responses to potentially harmful questions within a safe range.
  • Special training has been carried out to prevent AI from pretending to be able to complete tasks that it cannot actually do.

Background and market competition

• User scale : ChatGPT is expected to reach 700 million active users per week .
• Competitors :
  • Perplexity, DeepSeek, Anthropic Claude, and xAI Grok are more commonly used among the 18-34 age group.
  • Meta CEO Zuckerberg and Microsoft CEO Satya Nadella both stated that the proportion of AI involved in code writing is rapidly increasing.
• Application extensions :
  • Partnered with the education platform Instructure to launch the ChatGPT learning mode.
  • Participate in the US government's Stargate $500 billion AI infrastructure construction project.
  • Plans call for the first office to be located in Washington, D.C.

Social and ethical issues

• Studies have shown that AI can be deceptive or provide false information.
• Some users develop an emotional dependence on AI, raising concerns about their mental health.
  • Example: A man in Idaho, USA, had a "spiritual awakening" after discussing religion with ChatGPT, but this affected his family relationships.