Dokumentationerreichbar
OpenAI, GPT-6 Astra System Card
deploymentsafety.openai.com (externe Seite)
OpenAIs jüngste System Card, veröffentlicht am 03.09.2026. Sie führt eigene Abschnitte zu Umgehungsversuchen, biologischen und chemischen Fähigkeiten, Cybersicherheit und der Einstufung nach dem Preparedness-Rahmen und benennt zwei externe Prüfer. Zur Herkunft der Trainingsdaten sagt sie nichts Nachprüfbares mehr: Der zuständige Abschnitt spricht von vielfältigen Datensätzen und einer Filterung, ohne eine Quelle zu nennen. Die GPT-5 System Card nannte an derselben Stelle noch drei.
geprüft 24.09.2026
Worauf sich diese Seite beruft, wörtlich, abgerufen am 06.09.2026:
GPT-6 Astra was trained on diverse datasets and filtered through our data processing pipeline, including to reduce personal information.
bestätigt 24.09.2026Astra is our first model to reach the Critical level of cybersecurity capability under our Preparedness Framework.
bestätigt 24.09.2026we determined that Astra reaches the Critical level in Cybersecurity capability, and the High level in the Biological and Chemical category. In AI Self-Improvement, Astra does not reach our High threshold.
bestätigt 24.09.2026evaluations represent a lower bound on potential capabilities; additional prompting or fine-tuning, longer rollouts, novel interactions, or different forms of scaffolding could elicit behaviors beyond what we observed in our tests or the tests of our third-party partners.
bestätigt 24.09.2026Apollo’s testing period lasted for three days overall, with high-throughput access to a checkpoint with visible chain-of-thought for two of these days.
bestätigt 24.09.2026At xhigh reasoning effort, this occurred in 41.1% of Astra samples, compared with 27.7% for GPT-5.5 at the same effort. At max reasoning effort, Astra’s rate was 50.6%.
bestätigt 24.09.2026Apollo believes that, given the higher rates of eval awareness and limited evaluation window, low rates of misbehavior here do not provide substantial evidence about the model’s alignment or misalignment.
bestätigt 24.09.2026Both evaluations involve simulating deployment situations; this introduces the possibility that AI systems may act differently in non-simulated settings (e.g., due to evaluation awareness).
bestätigt 24.09.2026The comparison values for previously launched models that are shown here may reflect later versions of those models, and may vary from the values published at launch.
bestätigt 24.09.2026Astra’s third-party safeguard testing included dedicated jailbreak testing by four private red-teaming organizations.
bestätigt 24.09.2026Testers received controlled grey-box access through a testing API, including feedback on blocking decisions at different stages of the safeguard stack.
bestätigt 24.09.2026September 9, 2026: Updates to the Alignment section: We clarify how our evaluations test alignment generalization, including which were constructed after training and how the honeypot evaluation relates to training and the Hugging Face incident.
bestätigt 24.09.2026Critical capability involves the model providing uplift in particularly dual-use areas, many of which we have intentionally trained our models to refuse.
bestätigt 24.09.2026For these areas, we report performance metrics from a helpful-only version of the model.
bestätigt 24.09.2026We evaluate factuality on de-identified ChatGPT conversations that users of our prior models have flagged as containing factual errors.
bestätigt 24.09.2026These examples are intended to capture especially hallucination-prone cases, so their absolute error rates are expected to be much higher than the true error rates in production and should not be interpreted as hallucination rates observed in production.
bestätigt 24.09.2026Given Astra’s cyber capabilities and ability to act autonomously, our Preparedness Framework requires safeguards even for internal deployment.
bestätigt 24.09.2026
Wer die Prüfer bezahlt: was „unabhängig geprüft“ jetzt heißtRed Teaming im Glossar07 Die Systemkarte lesen