Get the weekly digest
Every policy change across 844 tracked documents, once a week. No account needed.
OpenAI's GPT-5 System Card was updated on July 18, 2026, with a modification to the document's linked content references. The change replaced references to three previously linked articles ('Separating signal from noise in coding evaluations', 'Introducing GeneBench-Pro', and a generic 'View all' link) with two new articles titled 'Why teens deserve access to safe AI' and 'GPT-Red: Unlocking Self-Improvement for Robustness'. The substantive text of the system card remains unchanged; only the curated set of related documents and safety research linked from the card has been updated.
This change does not affect how users interact with GPT-5 or what the model is permitted to do. The substantive content of the system card, including the model's capabilities and risk assessment for biological harm, remains unchanged. OpenAI has updated the curated set of related research articles linked from the card, replacing older safety research with newer materials on teen access and model robustness.
The GPT-5 System Card is OpenAI's primary public disclosure of the model's capabilities and safety assessment. Updates to this document are significant for transparency purposes. However, this particular change is limited to curating the linked research references; the substantive risk assessment and precautionary approach statements remain identical, so the material governance or safety disclosure position is unchanged.
Updated curated articles linked from the system card to feature newer safety research on teen access and model robustness.
This change record describes what was added, removed, or modified in the document. Analysis reflects what the updated agreement states or permits. It does not constitute a legal determination about enforceability. Applicability may vary by jurisdiction. Methodology
This is a documentation update with no operational or governance implications for organizations using GPT-5. The core safety assessment and risk disclosures stated in the system card remain identical. The change involves replacing linked references to research articles, which is a content curation modification rather than a substantive policy or capability change. No regulatory or compliance action is triggered.
Full compliance analysis
Regulatory exposure, obligation analysis, escalation trigger, board language, and recommended action.
Analyst $49/moConductAtlas provides verified policy intelligence sourced directly from platform documents. All analysis is intended to support, not replace, legal and compliance review. Record CA-C-003791.
OpenAI updated the GPT-5.5 System Card on April 24, 2026, to include safeguards information. In an update detected on July …
OpenAI removed the audio article feature reference from its Frontier Governance Framework post on July 18, 2026. The updated version …
OpenAI modified two sentences in its Enterprise Privacy policy as detected on July 18, 2026. The first change was purely …
H.R. 8094 would make the FTC the referee for AI model disclosure. It also names system cards as a way to comply, which turns a voluntary in…
The bill does not regulate most AI startups directly. But it changes the companies they depend on. Here is what the first federal AI law wo…
Ad personalization controls removed. Contact scanning added. Advertiser data partnerships quietly dropped. A timeline of every change.
Get alerted when this policy changes again — including what changed and why it matters.