{"total":62,"limit":100,"offset":0,"rows":[{"standard":"eu_gpai_code","item_id":"Appendix 3.5","item_title":"Independent external model evaluations","element_id":"C14","relation":"requires","quote":"Signatories will ensure that adequately qualified independent external evaluators conduct model evaluations pursuant to this Appendix 3, with regards to the systemic risk, unless: (1) the model is a similarly safe or safer model pursuant to Appendix 2.2; or (2) Signatories fail to appoint adequately qualified independent external evaluators, despite using early search efforts (such as through a public call open for 20 business days) and promptly notifying identified evaluators, in which case Signatories will take into account the potential additional uncertainty arising from the absence of independent external evaluations (pursuant to this Appendix 3.5) when determining whether the systemic risks stemming from the model are acceptable (pursuant to Commitment 4).","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Appendix 3.5","item_title":"Independent external model evaluations","element_id":"G4","relation":"requires","quote":"Signatories will ensure that adequately qualified independent external evaluators conduct model evaluations pursuant to this Appendix 3, with regards to the systemic risk, unless: (1) the model is a similarly safe or safer model pursuant to Appendix 2.2; or (2) Signatories fail to appoint adequately qualified independent external evaluators, despite using early search efforts (such as through a public call open for 20 business days) and promptly notifying identified evaluators, in which case Signatories will take into account the potential additional uncertainty arising from the absence of independent external evaluations (pursuant to this Appendix 3.5) when determining whether the systemic risks stemming from the model are acceptable (pursuant to Commitment 4).","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Appendix 4.2","item_title":"Protection of unreleased model parameters","element_id":"C08","relation":"requires","quote":"Signatories will protect unreleased model parameters by implementing security mitigations that achieve the following mitigation objectives: (1) accountability over all copies of stored model parameters across all devices and locations, through (a) a secure internal registry of all devices and locations where model parameters are stored; (2) prevention of unauthorised copying of model parameters to unmanaged devices, through (a) access management on all devices storing model parameters, with alerts in case of copying to unmanaged devices; (3) prevention of unauthorised access to model parameters during transport and at rest, through (a) ensuring model parameters are always encrypted during transportation and storage as appropriate, including encryption with at least 256 -bit security and with en cryption keys stored securely on a Trusted Platform Module (TPM); (4) prevention of unauthorised access to model parameters during temporary storage, through (a) ensuring model parameters are only decrypted for legitimate use to non -persistent memory; 42 (5) prevention of unauthorised access to model parameters during use, through (a) implementing confidential computing as appropriate, using hardware -based, and attested trusted execution environments; and (6) prevention of unauthorised physical access to systems hosting model parameters, through (a) restricting physical access to data centres and other sensitive working environments to required personnel only, along with regular inspections of such sites for un authorised personnel or devices.","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Appendix 4.3","item_title":"Hardening interface-access to unreleased model parameters","element_id":"C08","relation":"requires","quote":"Signatories will harden interface -access to unreleased model parameters while in use, by implementing security mitigations that achieve the following mitigation objectives: (1) prevention of unnecessary interface -access to model parameters, through (a) explicitly authorising only required software and persons for access to model parameters, enforced through multi-factor authentication mechanisms, and checked on a regular basis of at least every six months; (2) reduction of the risk of vulnerability exploitation or data leakage, through (a) thorough review of any software interfaces with access to model parameters by a security team to identify vulnerabilities or data leakage, and/or automated security reviews of any software interface code at least to the same standard as the highest level of automated security review used for other sensitive code; (3) reduction of the risk of model parameter exfiltration, through (a) hardening interfaces with access to model parameters, using methods such as output rate limiting; and (4) reduction of the risk of insider threats or compromised accounts, through (a) limiting the number of people who have non-hardened interface-access to model parameters.","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Appendix 4.4","item_title":"Insider threats","element_id":"C04","relation":"requires","quote":"(3) reduction of the risk of model self-exfiltration, through (a) sandboxes around models, such as virtual machines and code execution isolation;","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Appendix 4.4","item_title":"Insider threats","element_id":"C08","relation":"requires","quote":"Signatories will protect against insider threats, including in the form of (self-)exfiltration or sabotage carried out by models, by implementing security mitigations that achieve the following mitigation objectives: (1) protection of model parameters from insider threats attempting to gain work-related access with the Signatory, through (a) background checks on employees and contractors that have or might reasonably obtain read or write access to unreleased model parameters or systems that manage the access to such parameters;","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Appendix 4.5","item_title":"Security assurance","element_id":"C13","relation":"requires","quote":"validation of network software integrity, through (a) competitive bug bounty programs to encourage public participation in security testing of public-facing endpoints as appropriate to mitigate systemic risks;","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Commitment 6","item_title":"Security mitigations","element_id":"C08","relation":"requires","quote":"Signatories commit to implementing an adequate level of cybersecurity protection for their models and their physical infrastructure along the entire model lifecycle, as specified in the Measures for this Commitment, to ensure the systemic risks stemming fr om their models that could arise from unauthorised releases, unauthorised access, and/or model theft are acceptable (pursuant to Commitment 4).","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Commitment 9","item_title":"Serious incident reporting","element_id":"G3","relation":"requires","quote":"Signatories commit to implementing appropriate processes and measures for keeping track of, documenting, and reporting to the AI Office and, as applicable, to national competent authorities, without undue delay relevant information about serious incidents along the entire model lifecycle and possible corrective measures to address them, as specified in the Measures of this Commitment.","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Measure 1.1","item_title":"Creating the Framework","element_id":"G1","relation":"requires","quote":"Signatories will create a state -of-the-art Framework, taking into account the models they are developing, making available on the market, and/or using.","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Measure 1.4","item_title":"Framework notifications","element_id":"G1","relation":"requires","quote":"Signatories will provide the AI Office with (unredacted) access to their Framework, and updates thereof, within five business days of either being confirmed.","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Measure 3.5","item_title":"Post-market monitoring","element_id":"C06","relation":"requires","quote":"Signatories will conduct appropriate post -market monitoring to gather information relevant to assessing whether the systemic risk could be determined to not be acceptable (pursuant to Measure 4.1) and to inform whether a Model Report update is necessary (pursuant to Measure 7.6).","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Measure 3.5","item_title":"Post-market monitoring","element_id":"C07","relation":"recommends","quote":"monitoring aspects of models that are relevant for assessing and mitigating systemic risk and are not transparent to third parties, e.g. hidden chains -of-thought for models for which the parameters are not publicly available for download.","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Measure 3.5","item_title":"Post-market monitoring","element_id":"C10","relation":"recommends","quote":"implementing privacy-preserving logging and metadata analysis techniques of the model’s inputs and outputs using, e.g. watermarks, metadata, and/or other at least state -of-the-art provenance techniques;","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Measure 3.5","item_title":"Post-market monitoring","element_id":"C11","relation":"recommends","quote":"implementing privacy-preserving logging and metadata analysis techniques of the model’s inputs and outputs using, e.g. watermarks, metadata, and/or other at least state -of-the-art provenance techniques;","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Measure 3.5","item_title":"Post-market monitoring","element_id":"C13","relation":"recommends","quote":"providing bug bounties; 14","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Measure 3.5","item_title":"Post-market monitoring","element_id":"C14","relation":"requires","quote":"To facilitate post -market monitoring, Signatories will provide an adequate number of independent external evaluators with adequate free access to: (1) the model’s most capable model version(s) with regard to the systemic risk that is made available on the market; (2) the chains-of-thought of the model version(s) in point (1), if available; and (3) the model version(s) corresponding to the model version(s) in point (1) with the fewest safety mitigations implemented with regard to the systemic risk (such as the helpful -only model version, if it exists) and, as available, its chains-of-thought; unless the model is considered a similarly safe or safer model with regard to the same systemic risk (pursuant to Appendix 2.2).","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Measure 4.1","item_title":"Systemic risk acceptance criteria and acceptance determination","element_id":"G1","relation":"requires","quote":"define appropriate systemic risk tiers that: (i) are defined in terms of model capabilities, and may additionally incorporate model propensities, risk estimates, and/or other suitable metrics; (ii) are measurable; and (iii) comprise at least one systemic risk tier that has not been reached by the model","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Measure 5.1","item_title":"Appropriate safety mitigations","element_id":"C01","relation":"recommends","quote":"monitoring and filtering the model’s inputs and/or outputs","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Measure 5.1","item_title":"Appropriate safety mitigations","element_id":"C02","relation":"recommends","quote":"changing the model behaviour in the interests of safety, such as fine -tuning the model to refuse certain requests or provide unhelpful responses","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Measure 5.1","item_title":"Appropriate safety mitigations","element_id":"C09","relation":"recommends","quote":"staging the access to the model, e.g. by limiting API access to vetted users, gradually expanding access based on post -market monitoring, and/or not making the model parameters publicly available for download initially","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Measure 7.4","item_title":"External reports","element_id":"G4","relation":"requires","quote":"Signatories will provide in the Model Report: (1) any available reports (e.g. via valid hyperlinks) from: (a) independent external evaluators involved in model evaluations pursuant to Appendix 3.5; and (b) security reviews undertaken by an independent external party pursuant to Appendix 4.5; to the extent that respects existing confidentiality (including commercial confidentiality) obligations and allows such external evaluators or parties to maintain control over the publication of their findings, without implicit endorsement by the Signatories of the content of such reports;","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Measure 9.1","item_title":"Methods for serious incident identification","element_id":"G3","relation":"requires","quote":"facilitate the reporting of relevant information about serious incidents by downstream modifiers, downstream providers, users, and other third parties to: (a) the Signatory; or (b) the AI Office and, as applicable, national competent authorities; by informing such third parties of direct reporting channels, if available, without prejudice to any of their reporting obligations under Article 73 AI Act.","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Measure 9.2","item_title":"Relevant information for serious incident tracking, documentation, and reporting","element_id":"G3","relation":"requires","quote":"Signatories will keep track of, document, and report to the AI Office and, as applicable, to national competent authorities, at least the following information to the best of their knowledge, redacted to the extent necessary to comply with other Union law applicable to such information: (1) the start and end dates of the serious incident, or best approximations thereof if the precise dates are unclear; (2) the resulting harm and the victim or affected group of the serious incident; (3) the chain of events that (directly or indirectly) led to the serious incident; (4) the model involved in the serious incident; (5) a description of material available setting out the model’s involvement in the serious incident; (6) what, if anything, the Signatory intends to do or has done in response to the serious incident; (7) what, if anything, the Signatory recommends the AI Office and, as applicable, national competent authorities to do in response to the serious incident; (8) a root cause analysis with a description of the model’s outputs that (directly or indirectly) led to the serious incident and the factors that contributed to their generation, including the inputs used and any failures or circumventions of systemic risk mitigations; and (9) any patterns detected during post -market monitoring (pursuant to Measure 3.5) that can reasonably be assumed to be connected to the serious incident, such as individual or aggregate data on near misses.","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"eu_gpai_code","item_id":"Measure 9.3","item_title":"Reporting timelines","element_id":"G3","relation":"requires","quote":"Signatories will provide the information in points (1) to (7) of Measure 9.2 in an initial report that is submitted to the AI Office and, as applicable, to national competent authorities, at the following points in time, save in exceptional circumstances, if the involvement of their model (directly or indirectly) led to:","url":"https://ec.europa.eu/newsroom/dae/redirection/document/118119","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"mitre_atlas","item_id":"AML.M0005","item_title":"Control Access to AI Models and Data at Rest","element_id":"C08","relation":"recommends","quote":"Establish access controls on internal model registries and limit internal access to production models.","url":"https://atlas.mitre.org/mitigations/AML.M0005","retrieved":"2026-09-28","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"mitre_atlas","item_id":"AML.M0012","item_title":"Encrypt Sensitive Information","element_id":"C08","relation":"recommends","quote":"Encrypt sensitive data such as AI models to protect against adversaries attempting to access sensitive data.","url":"https://atlas.mitre.org/mitigations/AML.M0012","retrieved":"2026-09-28","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"mitre_atlas","item_id":"AML.M0015","item_title":"Adversarial Input Detection","element_id":"C01","relation":"recommends","quote":"Detect and block adversarial inputs or atypical queries that deviate from known benign behavior, exhibit behavior patterns observed in previous attacks or that come from potentially malicious IPs.","url":"https://atlas.mitre.org/mitigations/AML.M0015","retrieved":"2026-09-28","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"mitre_atlas","item_id":"AML.M0019","item_title":"Control Access to AI Models and Data in Production","element_id":"C06","relation":"requires","quote":"Require authentication for API endpoints and monitor production model queries to ensure compliance with usage policies and to prevent model misuse.","url":"https://atlas.mitre.org/mitigations/AML.M0019","retrieved":"2026-09-28","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"mitre_atlas","item_id":"AML.M0019","item_title":"Control Access to AI Models and Data in Production","element_id":"C09","relation":"requires","quote":"Require users to verify their identities before accessing a production model.","url":"https://atlas.mitre.org/mitigations/AML.M0019","retrieved":"2026-09-28","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"mitre_atlas","item_id":"AML.M0020","item_title":"Generative AI Guardrails","element_id":"C01","relation":"names","quote":"Guardrails can take the form of validators such as filters, rule-based logic, or regular expressions, as well as AI-based approaches, such as classifiers and utilizing LLMs, or named entity recognition (NER) to evaluate the safety of the prompt or response.","url":"https://atlas.mitre.org/mitigations/AML.M0020","retrieved":"2026-09-28","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"mitre_atlas","item_id":"AML.M0022","item_title":"Generative AI Model Alignment","element_id":"C02","relation":"recommends","quote":"When training or fine-tuning a generative AI model it is important to utilize techniques that improve model alignment with safety, security, and content policies.","url":"https://atlas.mitre.org/mitigations/AML.M0022","retrieved":"2026-09-28","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"mitre_atlas","item_id":"AML.M0024","item_title":"AI Telemetry Logging","element_id":"C11","relation":"recommends","quote":"Implement logging of inputs and outputs of deployed AI models.","url":"https://atlas.mitre.org/mitigations/AML.M0024","retrieved":"2026-09-28","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"mitre_atlas","item_id":"AML.M0026","item_title":"Privileged AI Agent Permissions Configuration","element_id":"C04","relation":"recommends","quote":"These controls include Role-Based Access Controls (RBAC), Attribute-Based Access Controls (ABAC), and the principle of least privilege so that the agent is only granted the necessary permissions to access tools and resources required to accomplish its designated task(s).","url":"https://atlas.mitre.org/mitigations/AML.M0026","retrieved":"2026-09-28","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"mitre_atlas","item_id":"AML.M0027","item_title":"Single-User AI Agent Permissions Configuration","element_id":"C04","relation":"recommends","quote":"Controls should also include the principle of least privilege and delegated access from the user account.","url":"https://atlas.mitre.org/mitigations/AML.M0027","retrieved":"2026-09-28","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"mitre_atlas","item_id":"AML.M0028","item_title":"AI Agent Tools Permissions Configuration","element_id":"C04","relation":"recommends","quote":"These controls include applying the principle of least privilege along with delegated access, where the tools receive the permissions, identities, and restrictions of the AI agent calling them.","url":"https://atlas.mitre.org/mitigations/AML.M0028","retrieved":"2026-09-28","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"mitre_atlas","item_id":"AML.M0029","item_title":"Human In-the-Loop for AI Agent Actions","element_id":"C05","relation":"recommends","quote":"Systems should require the user or another human stakeholder to approve AI agent actions before the agent takes them.","url":"https://atlas.mitre.org/mitigations/AML.M0029","retrieved":"2026-09-28","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"mitre_atlas","item_id":"AML.M0030","item_title":"Restrict AI Agent Tool Invocation on Untrusted Data","element_id":"C03","relation":"recommends","quote":"It is recommended that tool invocation be restricted or limited when untrusted data enters the LLM's context.","url":"https://atlas.mitre.org/mitigations/AML.M0030","retrieved":"2026-09-28","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"mitre_atlas","item_id":"AML.M0030","item_title":"Restrict AI Agent Tool Invocation on Untrusted Data","element_id":"C04","relation":"recommends","quote":"Consider blocking the automatic invocation of tools or requiring user confirmation once untrusted data enters the LLM's context.","url":"https://atlas.mitre.org/mitigations/AML.M0030","retrieved":"2026-09-28","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"mitre_atlas","item_id":"AML.M0030","item_title":"Restrict AI Agent Tool Invocation on Untrusted Data","element_id":"C05","relation":"recommends","quote":"For high consequence actions, consider always requiring user confirmation.","url":"https://atlas.mitre.org/mitigations/AML.M0030","retrieved":"2026-09-28","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"mitre_atlas","item_id":"AML.M0031","item_title":"Memory Hardening","element_id":"C04","relation":"recommends","quote":"This may be implemented using a combination of strategies including restricting an agent's ability to store memories by requiring external authentication and validation for memory updates, performing semantic integrity checks on retrieved memories before agents execute actions, and implementing controls for monitoring of memory and remediation processes for poisoned memory.","url":"https://atlas.mitre.org/mitigations/AML.M0031","retrieved":"2026-09-28","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"mitre_atlas","item_id":"AML.M0032","item_title":"Segmentation of AI Agent Components","element_id":"C04","relation":"recommends","quote":"Define security boundaries around agentic tools and data sources with methods such as API access, container isolation, code execution sandboxing, and rate limiting of tool invocation.","url":"https://atlas.mitre.org/mitigations/AML.M0032","retrieved":"2026-09-28","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"mitre_atlas","item_id":"AML.M0033","item_title":"Input and Output Validation for AI Agent Components","element_id":"C01","relation":"recommends","quote":"Implement validation on inputs and outputs for the tools and data sources used by AI agents.","url":"https://atlas.mitre.org/mitigations/AML.M0033","retrieved":"2026-09-28","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"mitre_atlas","item_id":"AML.M0033","item_title":"Input and Output Validation for AI Agent Components","element_id":"C03","relation":"recommends","quote":"Validation includes enforcing a common data format, schema validation, checks for sensitive or prohibited information leakage, and data sanitization to remove potential injections or unsafe code.","url":"https://atlas.mitre.org/mitigations/AML.M0033","retrieved":"2026-09-28","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"mitre_atlas","item_id":"AML.M0034","item_title":"Deepfake Detection","element_id":"C01","relation":"recommends","quote":"Apply deepfake detection algorithms against any untrusted or user-provided data, especially in impactful applications such as biometric verification, to block generated content.","url":"https://atlas.mitre.org/mitigations/AML.M0034","retrieved":"2026-09-28","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"nist_ai_rmf","item_id":"MANAGE 4.3","item_title":"Incidents and errors are communicated to relevant AI actors, including affected communities. Processes for tracking, responding to, and recovering from incidents and errors are followed and documented.","element_id":"G3","relation":"requires","quote":"Incidents and errors are communicated to relevant AI actors, including affected communities.","url":"https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.100-1.pdf","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"owasp_llm","item_id":"LLM01","item_title":"Prompt Injection","element_id":"C01","relation":"recommends","quote":"3. Implement input and output filtering Define sensitive categories and construct rules for identifying and handling such content.","url":"https://genai.owasp.org/llmrisk/llm01-prompt-injection/","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"owasp_llm","item_id":"LLM01","item_title":"Prompt Injection","element_id":"C03","relation":"recommends","quote":"However, the following measures can mitigate the impact of prompt injections: 1. Constrain model behavior Provide specific instructions about the model’s role, capabilities, and limitations within the system prompt.","url":"https://genai.owasp.org/llmrisk/llm01-prompt-injection/","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"owasp_llm","item_id":"LLM01","item_title":"Prompt Injection","element_id":"C04","relation":"recommends","quote":"4. Enforce privilege control and least privilege access Provide the application with its own API tokens for extensible functionality, and handle these functions in code rather than providing them to the model.","url":"https://genai.owasp.org/llmrisk/llm01-prompt-injection/","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"owasp_llm","item_id":"LLM01","item_title":"Prompt Injection","element_id":"C05","relation":"recommends","quote":"5. Require human approval for high-risk actions Implement human-in-the-loop controls for privileged operations to prevent unauthorized actions.","url":"https://genai.owasp.org/llmrisk/llm01-prompt-injection/","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"owasp_llm","item_id":"LLM05","item_title":"Improper Output Handling","element_id":"C11","relation":"recommends","quote":"Implement robust logging and monitoring systems to detect unusual patterns in LLM outputs that might indicate exploitation attempts.","url":"https://genai.owasp.org/llmrisk/llm052025-improper-output-handling/","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"owasp_llm","item_id":"LLM06","item_title":"Excessive Agency","element_id":"C04","relation":"recommends","quote":"Limit the extensions that LLM agents are allowed to call to only the minimum necessary.","url":"https://genai.owasp.org/llmrisk/llm062025-excessive-agency/","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"owasp_llm","item_id":"LLM06","item_title":"Excessive Agency","element_id":"C05","relation":"recommends","quote":"Utilise human-in-the-loop control to require a human to approve high-impact actions before they are taken.","url":"https://genai.owasp.org/llmrisk/llm062025-excessive-agency/","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"owasp_llm","item_id":"LLM06","item_title":"Excessive Agency","element_id":"C11","relation":"recommends","quote":"Log and monitor the activity of LLM extensions and downstream systems to identify where undesirable actions are taking place, and respond accordingly.","url":"https://genai.owasp.org/llmrisk/llm062025-excessive-agency/","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"owasp_llm","item_id":"LLM07","item_title":"System Prompt Leakage","element_id":"C01","relation":"recommends","quote":"An independent system that can inspect the output to determine if the model is in compliance with expectations is preferable to system prompt instructions.","url":"https://genai.owasp.org/llmrisk/llm072025-system-prompt-leakage/","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"owasp_llm","item_id":"LLM07","item_title":"System Prompt Leakage","element_id":"C02","relation":"names","quote":"While training particular behavior into a model can be effective, such as training it not to reveal its system prompt, it is not a guarantee that the model will always adhere to this.","url":"https://genai.owasp.org/llmrisk/llm072025-system-prompt-leakage/","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"owasp_llm","item_id":"LLM07","item_title":"System Prompt Leakage","element_id":"C04","relation":"recommends","quote":"In cases where an agent is performing tasks, if those tasks require different levels of access, then multiple agents should be used, each configured with the least privileges needed to perform the desired tasks.","url":"https://genai.owasp.org/llmrisk/llm072025-system-prompt-leakage/","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"owasp_llm","item_id":"LLM08","item_title":"Vector and Embedding Weaknesses","element_id":"C11","relation":"recommends","quote":"Maintain detailed immutable logs of retrieval activities to detect and respond promptly to suspicious behavior.","url":"https://genai.owasp.org/llmrisk/llm082025-vector-and-embedding-weaknesses/","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"owasp_llm","item_id":"LLM10","item_title":"Unbounded Consumption","element_id":"C04","relation":"recommends","quote":"Restrict the LLM’s access to network resources, internal services, and APIs.","url":"https://genai.owasp.org/llmrisk/llm102025-unbounded-consumption/","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"owasp_llm","item_id":"LLM10","item_title":"Unbounded Consumption","element_id":"C08","relation":"recommends","quote":"Implement strong access controls, including role-based access control (RBAC) and the principle of least privilege, to limit unauthorized access to LLM model repositories and training environments.","url":"https://genai.owasp.org/llmrisk/llm102025-unbounded-consumption/","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"owasp_llm","item_id":"LLM10","item_title":"Unbounded Consumption","element_id":"C10","relation":"recommends","quote":"Implement watermarking frameworks to embed and detect unauthorized use of LLM outputs.","url":"https://genai.owasp.org/llmrisk/llm102025-unbounded-consumption/","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"},{"standard":"owasp_llm","item_id":"LLM10","item_title":"Unbounded Consumption","element_id":"C11","relation":"recommends","quote":"Continuously monitor resource usage and implement logging to detect and respond to unusual patterns of resource consumption.","url":"https://genai.owasp.org/llmrisk/llm102025-unbounded-consumption/","retrieved":"2026-10-10","status":"proposed","confirmed_by":null,"method":"model deepseek-chat prompt crosswalk-2026-10-10.2"}]}