OpenAI is developing new open-weight models in conjunction with GPT-5 to enhance SEO performance.

OpenAI is not solely focused on the development of GPT-5; it is also preparing to release new open-source weights, aligning with its commitment to openness. Recently, two new open-source models, “gpt-oss-20b” and “gpt-oss-120b,” were discovered on HuggingFace, a prominent AI model hosting platform. These models have appeared on HuggingFace in advance of their public release, indicating that OpenAI is in the process of sharing model weights with partner organisations. This sharing is a standard practice in preparation for broader distribution. As OpenAI continues to make strides in this area, further references to these open-weight models are anticipated in the near future.

In addition to its advancements in open-source models, the cybersecurity landscape has seen a significant rise in malware targeting password stores, with incidents surging threefold. Attackers have employed stealthy tactics, executing what are termed “Perfect Heist” scenarios to infiltrate and exploit critical systems. This alarming trend highlights the importance of understanding and defending against prevalent attack techniques. For those interested in bolstering their cybersecurity measures, discovering the top ten MITRE ATT&CK techniques responsible for 93% of attacks could provide valuable insights. Mayank Parmar, a technology entrepreneur pursuing an MBA, covers these developments at BleepingComputer, with a particular focus on Microsoft and Windows-related stories.