Examples
Examples from published documents
- California SB 53 deems a developer compliant if the required information is published "as part of a larger document, including a system card or model card" (California SB 53, 22757.12(c)(3))
- OpenAI documents results "in a Safety and Security Model Report (referred to as 'Transparency Reports' under the TFAIA)" (OpenAI Frontier Governance Framework, §4, p.16)
- Meta "will publish a preparedness report for each closed or open Frontier AI release" (Meta Advanced AI Scaling Framework v2, §2.2.1, p.7)
Counter-examples
Not to be confused with
- A safety framework is the standing policy that reports are published under; it is not itself a per-model or periodic disclosure.
- An incident report is triggered by an event, such as a critical safety incident, rather than by a release or a reporting cadence.
Related terms
Related AI safety terms
Terms defined under this element that are not NIKOLAI elements themselves, each as the cited document uses it. Every term is also listed in the AI safety glossary.
- Model card
- A model card is a document published with an AI model that summarises its capabilities, known limitations, mitigations and safety results. Google DeepMind describes its model cards as essential information on Gemini models.
- Source: Google DeepMind, Gemini 3.7 Flash model card
- System card
- A system card is a launch document describing a model as deployed, including its safeguards, evaluations and risks. California SB 53 accepts a system card or model card as its transparency report if it holds the required information.
- Source: California SB 53
- Model Report
- Model Report, short for Safety and Security Model Report, is the EU Code of Practice document in which a signatory reports its model and systemic risk assessment to the AI Office before placing the model on the market.
- Source: EU GPAI Code of Practice, Safety and Security Chapter
- Preparedness report
- Preparedness report is Meta's per-release safety disclosure under its Advanced AI Scaling Framework, published for each closed or open frontier AI release in connection with the deployment.
- Source: Meta Advanced AI Scaling Framework v2
Source of record
Where this definition comes from
California SB 53, 22757.12(c)(1), (c)(3)
“"Before, or concurrently with, deploying a new frontier model or a substantially modified version of an existing frontier model, a frontier developer shall clearly and conspicuously publish on its internet website a transparency report"; "A frontier developer that publishes the information described in paragraph (1) or (2) as part of a larger document, including a system card or model card, shall be deemed in compliance with the applicable paragraph."”
https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202520260SB53EU GPAI Code of Practice, Safety and Security Chapter, Commitment 7
“"Signatories commit to reporting to the AI Office information about their model and their systemic risk assessment and mitigation processes and measures by creating a Safety and Security Model Report (“Model Report”) before placing a model on the market"”
https://ec.europa.eu/newsroom/dae/redirection/document/118119
Crosswalk
How named organisations use this concept
| Organisation | Their term, as published | Match & verification | Source |
|---|---|---|---|
| AnthropicShadow mapping Responsible Scaling Policy v3.4 / Risk Report | “"A Risk Report will cover the risks of Anthropic's models and actions as of a given coverage date"; "We will publish a Risk Report every 3-6 months." (§3.1, p.11)” A Risk Report is developer-wide and periodic, covering all in-scope models, unlike a per-model card. | exactEQ confidence: high | Anthropic Responsible Scaling Policy v3.4, effective 8 Jul 2026 |
| OpenAIShadow mapping Frontier Governance Framework | “"results of our systemic risk assessment and mitigation processes and measures are documented in a Safety and Security Model Report (referred to as 'Transparency Reports' under the TFAIA)" (§4, p.16).” OpenAI equates its EU-style Model Report with the SB 53 transparency report; it also publishes system cards at launch (§4). | exactEQ confidence: high | OpenAI Frontier Governance Framework |
| Google DeepMindShadow mapping Gemini model cards | “"Model cards are intended to provide essential information on Gemini models, including known limitations, mitigation approaches, and safety performance. Model cards may be updated from time to time" (card preamble, p.2).” Google DeepMind also publishes separate per-model FSF reports, so its model card is one of two related per-model documents. | closeCL confidence: high | Google DeepMind, Gemini 3.7 Flash model card |
| MetaShadow mapping Advanced AI Scaling Framework v2 | “"We will publish a preparedness report for each closed or open Frontier AI release in a timely manner in connection with the deployment" (§2.2.1, p.7).” | exactEQ confidence: high | Meta Advanced AI Scaling Framework v2 |
| xAIShadow mapping Grok 4.6 model card | “Document titled "Model Card: Grok 4.6", with a cover revision date and a dated changelog (cover; p.1).” | closeCL confidence: medium | Grok 4.6 model card |
| EUShadow mapping Safety and Security chapter | “"Safety and Security Model Report (“Model Report”)" created "before placing a model on the market" (Commitment 7); updates carry "a changelog, describing how and why the Model Report has been updated, along with a version number and the date of change" (Measure 7.6).” | exactEQ confidence: high | EU GPAI Code of Practice, Safety and Security Chapter |
| California SB 53Shadow mapping SB 53 statute text | “Transparency report published "Before, or concurrently with, deploying a new frontier model" (22757.12(c)(1)); a "system card or model card" that contains the required information is deemed compliant (22757.12(c)(3)).” SB 53 does not define "transparency report"; its contents are set by 22757.12(c)(1)-(2). | exactEQ confidence: high | California SB 53 |
What do these codes mean?
- exact
- The source term is equivalent to this element
- close
- The source term is close but not equivalent to this element
- broad
- The source term is broader than this element
- narrow
- The source term is narrower than this element
- none
- No mapping claim — used for false-friend and declared-but-undefined rows
- EQ
- Equivalent
- CL
- Close
- BR
- Source is broader than the element
- NR
- Source is narrower than the element
- FF
- False friend — same or similar label, different meaning
- DU
- Declared but undefined by the source
- UV
- Unverified
Related, not mapped
Pointers that are not crosswalk claims
These sources mention this concept but do not define or map it clearly enough to count as a crosswalk row — noted here so the research is visible without overstating it as a mapping.
- Amazon
Amazon's framework commits to publishing evaluation information with launches in model documentation such as model service cards (s.4, p.5; CASRAI's paraphrase, not a quotation).
Amazon's Frontier Model Safety Framework
Divergence
Where sources materially disagree
The names do not map one to one. OpenAI treats its Safety and Security Model Report as the SB 53 transparency report and publishes system cards separately; SB 53 lets a system card or model card serve as the transparency report; Anthropic's Risk Report covers all in-scope models at once rather than one model. "Model card" and "system card" are often used for the same document, but some developers distinguish the model from the deployed system.
Gap
SB 53 leaves "transparency report" undefined beyond its required contents, and no source defines a shared vocabulary of document types. A reader comparing two disclosures has to work out first which kind each one is.
