OpenAI Launches Open-Weight Language Models and Enhances ChatGPT Focus, While Google’s AI Tool Uncovers Security Flaws

OpenAI has launched two new language models—GPT-OSS-120B and GPT-OSS-20B—designed for advanced reasoning and efficient tool use, with the added advantage of running on consumer hardware. These models are fully customizable, support structured outputs, and come with an Apache 2.0 license. In addition, ChatGPT now prioritizes helping users achieve their goals over tracking engagement metrics. Meanwhile, Google’s DeepMind and Project Zero unveiled “Big Sleep,” an AI agent that has already discovered 20 security flaws in open-source software, marking a significant step forward in AI-powered cybersecurity research.

Table of Contents

• New Releases from OpenAI
• Launch of GPT-OSS-120B and GPT-OSS-20B
• Refinement of ChatGPT
• Google’s AI Bug Hunter: Big Sleep
• Conclusion

New Releases from OpenAI

Launch of GPT-OSS-120B and GPT-OSS-20B

OpenAI has just released two robust language models optimized for:

• Challenging reasoning tasks
• Efficient tool use

Key Features:

• Cost-effective deployment: These models can run on consumer hardware.
• Fully customizable: Supports structured outputs.
• Licensing: Available under the Apache 2.0 license.

Refinement of ChatGPT

OpenAI is making a noteworthy shift in how it evaluates user interaction:

• Focus on user progress: Moving away from traditional engagement metrics like time spent on the platform.
• Goal: Help users achieve their intended tasks.

This change is especially crucial as AI’s emotional impact becomes more recognized, particularly for vulnerable individuals. OpenAI’s mission is to empower users to navigate personal challenges rather than dictate their choices.

Google’s AI Bug Hunter: Big Sleep

Google’s DeepMind and Project Zero have developed a new vulnerability research agent, named Big Sleep.

Achievements:

• Identified: 20 confirmed security flaws in common open-source software.

This tool highlights the growing role of AI in cybersecurity, enhancing protection for both users and organizations alike.

Conclusion

That wraps up our news segment for today! We hope these updates keep you well-informed about the latest in Artificial Intelligence.

Stay connected, and tune in for more news as it develops. Until next time, take care!