Skip to content

AI explained in plain English. Read the latest articles

News Beginner

OpenAI launches GPT-6.2 Astra and Sol at highest risk level

OpenAI released GPT-6.2 Astra and Sol on 17 November 2026, and says both models score at the highest risk level for cybersecurity.

Digital network of light points as a symbol of a new AI model from OpenAI

OpenAI released GPT-6.2 Astra together with GPT-6.2 Sol on 17 November 2026. According to OpenAI, Astra is the first externally released model to reach the High threshold for AI Self-Improvement. OpenAI also rates both new models as Critical for cybersecurity. That is OpenAI’s own assessment and has not been independently checked.

What exactly is new?

OpenAI describes GPT-6.2 Astra and GPT-6.2 Sol in its own system card. It says both models were launched on 17 November 2026. The most striking point is the risk rating.

According to OpenAI, GPT-6.2 Astra reaches the High threshold for AI Self-Improvement. OpenAI says this is the first externally released model at this level. This is OpenAI’s own classification. The sources found do not show any independent check of this assessment.

GPT-6.2 Astra and GPT-6.2 Sol are also both given a Critical level for cybersecurity. That is the highest risk category mentioned in the coverage. Here too, this is OpenAI’s own classification.

A report saying OpenAI would have skipped GPT-6.1 is not correct. GPT-6.1 Sol is mentioned as the reference model used by OpenAI and Microsoft for comparison. That appears in the AI Daily Digest and is also mentioned in a news roundup by AI Weekly.

What does this mean for you?

For ordinary workers, nothing visible changes today. The sources do not confirm that GPT-6.2 Astra or GPT-6.2 Sol are already available in your tools. The news mainly signals that AI providers themselves are warning about risks around cyberattacks and self-learning systems.

If you use AI at work, it is wise to know which model your tool uses. Check the settings or documentation of your AI tool. Do not share sensitive company data without permission, especially if you do not know which model is behind it.

For entrepreneurs, this is a good moment to get the basics in order. Make sure your IT security and updates are up to date. Also check whether AI tools in your company can automatically switch to a new model. If they can, you will want to know in advance what that means for your data, your security and your agreements with customers.

Also review your AI policy. It should set out which AI tools your organisation may use, which data must not be shared and who is responsible for oversight.

When can you use it in the Netherlands?

Nothing has yet been confirmed about availability in the Netherlands or the price. The sources found do not confirm whether GPT-6.2 Astra and GPT-6.2 Sol can already be used in ChatGPT or via the API. It has also not been confirmed what use in the Netherlands or Belgium will cost.

If you want to use this model, first check OpenAI’s announcement or system card. Pay special attention to availability, price, terms and which version your AI tool is actually using.

If you want to prepare your organisation for this kind of model switch, read more about creating an AI policy.

Was this article helpful?

New articles every day

Follow AIzine and never miss an article

We publish three new articles about AI every day. Follow us on Instagram or Facebook to see them in your feed.