Hot details continue to come after the last minute development that occupies the agenda. According to the latest information obtained; OpenAI announced its new artificial intelligence system called GPT-Red. Legal investigations and investigations by the competent authorities on the subject continue to deepen. We will continue to report new developments as they occur.
Information from official sources regarding the issue, which has a wide repercussion on Türkiye's agenda, has become clear. According to the statements made; According to the news in Webtekno, GPT-Red is an internal artificial intelligence model developed entirely for OpenAI's security teams. Its mission is to test existing and future GPT models like an attacker and reveal their weak points. The company says that in this way, new models can be made much more secure before they reach real users. Detailed analysis and official statements of the issue are closely followed by our editorial team in line with the developments.
Hot details continue to come after the last minute development that occupies the agenda. According to the latest information obtained; According to the information shared by OpenAI, GPT-Red adopts an automatic "red teaming" approach, unlike classical security tests. In other words, the model tries to deceive artificial intelligence systems, bypass security measures and perform attacks such as command injection. Each successful scenario turns into new training data. While the investigation process carried out by the official authorities regarding the incident continues, you can instantly follow the most up-to-date information on our website.
Hot details continue to come after the last minute development that occupies the agenda. According to the latest information obtained; The method used by the company is called "self-play". While GPT-Red generates attacks, the defense model on the other side tries to prevent these attacks. Both parties continue to improve throughout the process. OpenAI states that this method enables the creation of many more scenarios in a short time compared to manual security tests. Legal investigations and investigations by the competent authorities on the subject continue to deepen. We will continue to report new developments as they occur.
According to the latest information about hot developments, a new phase has been entered in the event that is closely followed by the public. According to the details given; OpenAI also announced that GPT-Red is not just an experimental project and shared the details with the public. According to the data shared by the company, attack examples created by GPT-Red were used during GPT-5.6's security training. Thus, it was reported that the model became approximately 6 times more resistant than the previous production model, especially against direct command injection attacks. Detailed analysis and official statements of the issue are closely followed by our editorial team in line with the developments.
Information from official sources regarding the issue, which has a wide repercussion on Türkiye's agenda, has become clear. According to the statements made; One of the striking points is that this increase in security was achieved without compromising the general capabilities of the model. According to OpenAI, while GPT-5.6 maintains its performance in normal usage scenarios, it has only become more resistant to malicious commands. While the investigation process carried out by the official authorities regarding the incident continues, you can instantly follow the most up-to-date information on our website.
Information from official sources regarding the issue, which has a wide repercussion on Türkiye's agenda, has become clear. According to the statements made; OpenAI specifically emphasizes that GPT-Red will not be accessible via ChatGPT or API. The main reason for this is that the model is deliberately trained with attack generation capabilities. The company states that it will use this technology only for its own security research and that users will indirectly benefit from more secure GPT models. Legal investigations and investigations by the competent authorities on the subject continue to deepen. We will continue to report new developments as they occur.
Hot details continue to come after the last minute development that occupies the agenda. According to the latest information obtained; OpenAI also announced that it will soon publish the scientific preprint, which will contain much more detailed technical information about GPT-Red. The company aims to make new generation artificial intelligence models both safer and more durable by using more advanced versions of GPT-Red in the future. Detailed analysis and official statements of the issue are closely followed by our editorial team in line with the developments.