{"id":"2951cacb0fa934280c51c80980e0f9e08391b1172f52a899c19c4af1e2c43e70","title":"IEEE Spectrum: Can We Trust AI Agents When Zero Trust Isn’t Really Zero","argument":"An IEEE Spectrum article, brought to readers by Queen’s University Belfast, argues that agentic systems raise a basic trust problem: deleting historic information or enforcing access policies does not, by itself, show that agents followed instructions or cannot recover information later. The article describes this as a context recovery attack and asks where proof of compliance comes from when a policy enforcement point (PEP) must itself be trusted.\n\nThe article reports that a Queen’s University Belfast in-house experiment found agentic orchestration brought the rate of de-authentication attacks as low as 2 to 3 seconds, leaving little time to respond. It also warns that giving an untrusted AI model an organization’s codebase for vibe coding can expose vulnerabilities across multiple surfaces. These examples support the article’s concern that, when agents act across connected systems, recovery from failure may extend beyond one task or organization.\n\nThe article argues that trust is harder to establish in stateless systems and under asynchronous autonomy, where agents’ actions and delays complicate traceability. It discusses AI-bill of material and distributed ledger technology as possible infrastructure for traceability and immutability, while leaving open how to detect horizontal anomalies and govern trust.","topicIds":["power-and-infrastructure","work-and-agency"],"domain":"work","claimType":"attributed reading","limits":"The article is brought to readers by Queen’s University Belfast, but the supplied text does not identify an individual author. The article describes the de-authentication result as an in-house experiment but does not give its setup, measurement method, or conditions. The article raises possible approaches, including AI-bill of material and distributed ledger technology, but the supplied text does not report results demonstrating their effectiveness.","speaker":{"name":"Vishal Sharma","roleAtSource":"Article author"},"source":{"title":"Can We Trust AI Agents When Zero Trust Isn’t Really Zero","publisher":"IEEE Spectrum","format":"Article","url":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","publishedAt":"2026-11-09T16:25:06+00:00","dateBasis":"Date stated by the publisher."},"evidence":[{"text":"The article argues that deletion checkpoints cannot establish that information is unrecoverable, because a later prompt could bring it back.","excerpt":"a future prompt could make the agents bring forward the information that was supposed to be deleted","url":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","relationship":"Source-stated"},{"text":"The article says an in-house Queen’s University Belfast experiment found agentic orchestration could carry out de-authentication attacks in as little as 2 to 3 seconds.","excerpt":"the swift rate to do de-authentication attacks could go as low as 2 to 3 seconds.","url":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","relationship":"Source-stated"},{"text":"The article warns that giving an untrusted entity an organization’s codebase for vibe coding can lead to vulnerabilities across multiple surfaces.","excerpt":"giving away the entire codebase to an untrusted entity outside the organization can lead to vulnerability sweeping across multiple surfaces.","url":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","relationship":"Source-stated"},{"text":"The article argues that a single policy enforcement point may not scale to all agentic interactions.","excerpt":"placing a single entity and expecting it to scale towards all agentic interactions is underestimating the problem space.","url":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","relationship":"Source-stated"}],"sourceMedia":{"status":"found","sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","checkedAt":"2026-11-09T21:16:40.862Z","items":[{"type":"image","url":"https://spectrum.ieee.org/media-library/collage-of-indoor-drone-flight-phone-controls-and-operator-at-computer-workstation.jpg?id=68578379&width=1200&height=600&coordinates=0%2C80%2C0%2C80","sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","attribution":"IEEE Spectrum","basis":"publisher-metadata","availability":"verified","verifiedAt":"2026-11-09T21:16:41.384Z"}],"note":"Media declared by the original publisher; availability checked separately from the editorial reading."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"What evidence could show that agents have deleted historic information and cannot recover it through later prompts?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"IEEE Spectrum: Can We Trust AI Agents When Zero Trust Isn’t Really Zero","summary":"An IEEE Spectrum article, brought to readers by Queen’s University Belfast, argues that agentic systems raise a basic trust problem: deleting historic information or enforcing access policies does not, by itself, show that agents followed instructions or cannot recover information later. The article describes this as a context recovery attack and asks where proof of compliance comes from when a policy enforcement point (PEP) must itself be trusted.\n\nThe article reports that a Queen’s University Belfast in-house experiment found agentic orchestration brought the rate of de-authentication attacks as low as 2 to 3 seconds, leaving little time to respond. It also warns that giving an untrusted AI model an organization’s codebase for vibe coding can expose vulnerabilities across multiple surfaces. These examples support the article’s concern that, when agents act across connected systems, recovery from failure may extend beyond one task or organization.\n\nThe article argues that trust is harder to establish in stateless systems and under asynchronous autonomy, where agents’ actions and delays complicate traceability. It discusses AI-bill of material and distributed ledger technology as possible infrastructure for traceability and immutability, while leaving open how to detect horizontal anomalies and govern trust.","question":"What evidence could show that agents have deleted historic information and cannot recover it through later prompts?","claims":[{"text":"The article argues that deletion checkpoints cannot establish that information is unrecoverable, because a later prompt could bring it back.","excerpt":"a future prompt could make the agents bring forward the information that was supposed to be deleted","support":"source-stated"},{"text":"The article says an in-house Queen’s University Belfast experiment found agentic orchestration could carry out de-authentication attacks in as little as 2 to 3 seconds.","excerpt":"the swift rate to do de-authentication attacks could go as low as 2 to 3 seconds.","support":"source-stated"},{"text":"The article warns that giving an untrusted entity an organization’s codebase for vibe coding can lead to vulnerabilities across multiple surfaces.","excerpt":"giving away the entire codebase to an untrusted entity outside the organization can lead to vulnerability sweeping across multiple surfaces.","support":"source-stated"},{"text":"The article argues that a single policy enforcement point may not scale to all agentic interactions.","excerpt":"placing a single entity and expecting it to scale towards all agentic interactions is underestimating the problem space.","support":"source-stated"}],"limitations":["The article is brought to readers by Queen’s University Belfast, but the supplied text does not identify an individual author.","The article describes the de-authentication result as an in-house experiment but does not give its setup, measurement method, or conditions.","The article raises possible approaches, including AI-bill of material and distributed ledger technology, but the supplied text does not report results demonstrating their effectiveness."],"mediaId":"4f2d229e2ac04f8a3d9366c918a446f1f745fee8d9cc033e94a07638ac3f8782","correctionReason":null,"id":"2951cacb0fa934280c51c80980e0f9e08391b1172f52a899c19c4af1e2c43e70","publicationId":"3cde7f25-965a-4456-8fb1-d8fd57729d65","version":1,"sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","sourceTitle":"Can We Trust AI Agents When Zero Trust Isn’t Really Zero","publisher":"IEEE Spectrum","publishedAt":"2026-11-09T16:25:06+00:00","originalPublished":{"value":"2026-11-09T16:25:06+00:00","precision":"instant","basis":"publisher_page"},"firstObservedAt":"2026-11-09T18:27:37.407Z","createdAt":"2026-11-10T00:01:09.505Z","historical":false,"entityIds":["ieee-spectrum"],"topicIds":["power-and-infrastructure","work-and-agency"],"media":[{"id":"4f2d229e2ac04f8a3d9366c918a446f1f745fee8d9cc033e94a07638ac3f8782","kind":"article","url":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-09T21:16:41.384Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."},{"id":"482ccfab586cfaf46252ec9a28049e2a8b31aa2215c65e1edb8823db8dec2597","kind":"image","url":"https://spectrum.ieee.org/media-library/collage-of-indoor-drone-flight-phone-controls-and-operator-at-computer-workstation.jpg?id=68578379&width=1200&height=600&coordinates=0%2C80%2C0%2C80","sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","label":"View source image","provenance":"publisher_page","verification":{"status":"verified","checkedAt":"2026-11-09T21:16:41.384Z","method":"publisher_association_and_head","contentType":"image/jpeg","cacheUntil":1794259301384},"recommendationReason":"Observed on the acquired publisher source; availability checked."}],"sourceMedia":{"status":"found","sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","checkedAt":"2026-11-09T21:16:40.862Z","items":[{"type":"image","url":"https://spectrum.ieee.org/media-library/collage-of-indoor-drone-flight-phone-controls-and-operator-at-computer-workstation.jpg?id=68578379&width=1200&height=600&coordinates=0%2C80%2C0%2C80","sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","attribution":"IEEE Spectrum","basis":"publisher-metadata","availability":"verified","verifiedAt":"2026-11-09T21:16:41.384Z"}],"note":"Media declared by the original publisher; availability checked separately from the editorial reading."},"editorialDecision":{"outcome":"accept","reasonCodes":["useful_perspective"],"rubricVersion":"stream-admission-v5","mediaId":"4f2d229e2ac04f8a3d9366c918a446f1f745fee8d9cc033e94a07638ac3f8782","confidence":0.6900000000000001,"provider":"typesafe","model":"jev-1.13.0","inputHash":"009ff0fd8fda0bd13742e525a034b056e1d33862bc98f08b16e8bcd33c183f30","decidedAt":"2026-11-09T21:16:43.685Z"},"provenance":{"textHash":"e5ed1d230de64fd7897412a5531f2fc6266494d6d7e89d54fca19748cbd30f09","observationId":"d95a4ecccaf2f1f1c050caeb0c686d10abe80aa8df5ed0712ae9dcae8067ce80","retrievedAt":"2026-11-09T21:16:40.862Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"dd2e1783-058d-46de-824c-a13e6e73bd0c"},"reviewer":{"model":"gpt-6-luna","runId":"ca3a1925-e846-47d1-8036-4dbd029a7caf"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-10T00:01:09.505Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Vishal Sharma","url":"https://spectrum.ieee.org/u/vishal-sharma2","kind":"Person"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Vishal Sharma","url":"https://spectrum.ieee.org/u/vishal-sharma2","kind":"Person","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Vishal Sharma","status":"stated","provenance":{"basis":"article-jsonld","sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","locator":"document/html[2]/head[0]/script[0]@163.author"},"assertionId":"bd7efb0b86de563d4ea42bb925973e1228e915af0f68b6748338580d7323d0bf:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T21:16:40.862Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"bd7efb0b86de563d4ea42bb925973e1228e915af0f68b6748338580d7323d0bf","sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","contentSha256":"a5a5a40a7297b60e4100433d07bb64265e57888491777e25bee87218043ad7bd","textHash":"e5ed1d230de64fd7897412a5531f2fc6266494d6d7e89d54fca19748cbd30f09","observedAt":"2026-11-09T21:16:40.862Z"}},{"actor":{"name":"IEEE Spectrum","url":null,"kind":"unknown","actorId":null},"role":"publisher","scope":{"kind":"publication","locator":null},"literalCredit":"IEEE Spectrum","status":"stated","provenance":{"basis":"article-publisher","sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","locator":"document/html[2]/head[0]/script[0]@163.publisher"},"assertionId":"bd7efb0b86de563d4ea42bb925973e1228e915af0f68b6748338580d7323d0bf:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T21:16:40.862Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"bd7efb0b86de563d4ea42bb925973e1228e915af0f68b6748338580d7323d0bf","sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","contentSha256":"a5a5a40a7297b60e4100433d07bb64265e57888491777e25bee87218043ad7bd","textHash":"e5ed1d230de64fd7897412a5531f2fc6266494d6d7e89d54fca19748cbd30f09","observedAt":"2026-11-09T21:16:40.862Z"}}],"participationCoverage":{"total":2,"returned":2,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":{"name":"IEEE Spectrum","url":null},"sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","snapshotId":"bd7efb0b86de563d4ea42bb925973e1228e915af0f68b6748338580d7323d0bf","contentSha256":"a5a5a40a7297b60e4100433d07bb64265e57888491777e25bee87218043ad7bd","textHash":"e5ed1d230de64fd7897412a5531f2fc6266494d6d7e89d54fca19748cbd30f09","observedAt":"2026-11-09T21:16:40.862Z","basis":"publisher-page-metadata","sourceLinks":13,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/2951cacb0fa934280c51c80980e0f9e08391b1172f52a899c19c4af1e2c43e70/source"}},"canonicalUrl":"https://escapemeridian.com/readings/ieee-spectrum-can-we-trust-ai-agents-when-zero-trust-isn-t-2951cacb","formats":{"html":"/readings/ieee-spectrum-can-we-trust-ai-agents-when-zero-trust-isn-t-2951cacb","markdown":"/readings/ieee-spectrum-can-we-trust-ai-agents-when-zero-trust-isn-t-2951cacb.md","json":"/readings/ieee-spectrum-can-we-trust-ai-agents-when-zero-trust-isn-t-2951cacb.json"},"attribution":"Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.","maintainedGuideLinks":[],"readerContext":{"version":"reader-context-v1","attribution":"Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.","terms":[{"id":"term-agentic","name":"Agentic AI","explanation":"AI systems that plan and carry out a series of steps toward a goal, often by using other software, rather than giving one answer.","referenceUrl":"https://developers.google.com/machine-learning/glossary#agentic"}],"entities":[{"id":"ieee-spectrum","name":"IEEE Spectrum","kind":"organization","context":"IEEE Spectrum reports on engineering practice, research, and the industries around it.","referenceUrls":["https://spectrum.ieee.org/","https://spectrum.ieee.org/feeds/feed.rss"],"path":"/entities/ieee-spectrum","json":"/entities/ieee-spectrum.json","markdown":"/entities/ieee-spectrum.md"},{"id":"time","name":"TIME","kind":"organization","context":"TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.","referenceUrls":["https://time.com/"],"path":"/entities/time","json":"/entities/time.json","markdown":"/entities/time.md"}]},"researchQuestions":[],"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"2951cacb0fa934280c51c80980e0f9e08391b1172f52a899c19c4af1e2c43e70","revision":"3cde7f25-965a-4456-8fb1-d8fd57729d65","revisionKind":"publication-id","title":"IEEE Spectrum: Can We Trust AI Agents When Zero Trust Isn’t Really Zero","url":"https://escapemeridian.com/readings/ieee-spectrum-can-we-trust-ai-agents-when-zero-trust-isn-t-2951cacb"},"source":{"type":"source","id":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","revision":"e5ed1d230de64fd7897412a5531f2fc6266494d6d7e89d54fca19748cbd30f09","revisionKind":"sha256-source-text","title":"Can We Trust AI Agents When Zero Trust Isn’t Really Zero","url":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture"},"status":"available","items":[],"next":null,"scope":"Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter."},"revisions":[{"id":"2951cacb0fa934280c51c80980e0f9e08391b1172f52a899c19c4af1e2c43e70","title":"IEEE Spectrum: Can We Trust AI Agents When Zero Trust Isn’t Really Zero","argument":"An IEEE Spectrum article, brought to readers by Queen’s University Belfast, argues that agentic systems raise a basic trust problem: deleting historic information or enforcing access policies does not, by itself, show that agents followed instructions or cannot recover information later. The article describes this as a context recovery attack and asks where proof of compliance comes from when a policy enforcement point (PEP) must itself be trusted.\n\nThe article reports that a Queen’s University Belfast in-house experiment found agentic orchestration brought the rate of de-authentication attacks as low as 2 to 3 seconds, leaving little time to respond. It also warns that giving an untrusted AI model an organization’s codebase for vibe coding can expose vulnerabilities across multiple surfaces. These examples support the article’s concern that, when agents act across connected systems, recovery from failure may extend beyond one task or organization.\n\nThe article argues that trust is harder to establish in stateless systems and under asynchronous autonomy, where agents’ actions and delays complicate traceability. It discusses AI-bill of material and distributed ledger technology as possible infrastructure for traceability and immutability, while leaving open how to detect horizontal anomalies and govern trust.","topicIds":["power-and-infrastructure","work-and-agency"],"domain":"work","claimType":"attributed reading","limits":"The article is brought to readers by Queen’s University Belfast, but the supplied text does not identify an individual author. The article describes the de-authentication result as an in-house experiment but does not give its setup, measurement method, or conditions. The article raises possible approaches, including AI-bill of material and distributed ledger technology, but the supplied text does not report results demonstrating their effectiveness.","speaker":{"name":"Vishal Sharma","roleAtSource":"Article author"},"source":{"title":"Can We Trust AI Agents When Zero Trust Isn’t Really Zero","publisher":"IEEE Spectrum","format":"Article","url":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","publishedAt":"2026-11-09T16:25:06+00:00","dateBasis":"Date stated by the publisher."},"evidence":[{"text":"The article argues that deletion checkpoints cannot establish that information is unrecoverable, because a later prompt could bring it back.","excerpt":"a future prompt could make the agents bring forward the information that was supposed to be deleted","url":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","relationship":"Source-stated"},{"text":"The article says an in-house Queen’s University Belfast experiment found agentic orchestration could carry out de-authentication attacks in as little as 2 to 3 seconds.","excerpt":"the swift rate to do de-authentication attacks could go as low as 2 to 3 seconds.","url":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","relationship":"Source-stated"},{"text":"The article warns that giving an untrusted entity an organization’s codebase for vibe coding can lead to vulnerabilities across multiple surfaces.","excerpt":"giving away the entire codebase to an untrusted entity outside the organization can lead to vulnerability sweeping across multiple surfaces.","url":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","relationship":"Source-stated"},{"text":"The article argues that a single policy enforcement point may not scale to all agentic interactions.","excerpt":"placing a single entity and expecting it to scale towards all agentic interactions is underestimating the problem space.","url":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","relationship":"Source-stated"}],"sourceMedia":{"status":"found","sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","checkedAt":"2026-11-09T21:16:40.862Z","items":[{"type":"image","url":"https://spectrum.ieee.org/media-library/collage-of-indoor-drone-flight-phone-controls-and-operator-at-computer-workstation.jpg?id=68578379&width=1200&height=600&coordinates=0%2C80%2C0%2C80","sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","attribution":"IEEE Spectrum","basis":"publisher-metadata","availability":"verified","verifiedAt":"2026-11-09T21:16:41.384Z"}],"note":"Media declared by the original publisher; availability checked separately from the editorial reading."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"What evidence could show that agents have deleted historic information and cannot recover it through later prompts?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"IEEE Spectrum: Can We Trust AI Agents When Zero Trust Isn’t Really Zero","summary":"An IEEE Spectrum article, brought to readers by Queen’s University Belfast, argues that agentic systems raise a basic trust problem: deleting historic information or enforcing access policies does not, by itself, show that agents followed instructions or cannot recover information later. The article describes this as a context recovery attack and asks where proof of compliance comes from when a policy enforcement point (PEP) must itself be trusted.\n\nThe article reports that a Queen’s University Belfast in-house experiment found agentic orchestration brought the rate of de-authentication attacks as low as 2 to 3 seconds, leaving little time to respond. It also warns that giving an untrusted AI model an organization’s codebase for vibe coding can expose vulnerabilities across multiple surfaces. These examples support the article’s concern that, when agents act across connected systems, recovery from failure may extend beyond one task or organization.\n\nThe article argues that trust is harder to establish in stateless systems and under asynchronous autonomy, where agents’ actions and delays complicate traceability. It discusses AI-bill of material and distributed ledger technology as possible infrastructure for traceability and immutability, while leaving open how to detect horizontal anomalies and govern trust.","question":"What evidence could show that agents have deleted historic information and cannot recover it through later prompts?","claims":[{"text":"The article argues that deletion checkpoints cannot establish that information is unrecoverable, because a later prompt could bring it back.","excerpt":"a future prompt could make the agents bring forward the information that was supposed to be deleted","support":"source-stated"},{"text":"The article says an in-house Queen’s University Belfast experiment found agentic orchestration could carry out de-authentication attacks in as little as 2 to 3 seconds.","excerpt":"the swift rate to do de-authentication attacks could go as low as 2 to 3 seconds.","support":"source-stated"},{"text":"The article warns that giving an untrusted entity an organization’s codebase for vibe coding can lead to vulnerabilities across multiple surfaces.","excerpt":"giving away the entire codebase to an untrusted entity outside the organization can lead to vulnerability sweeping across multiple surfaces.","support":"source-stated"},{"text":"The article argues that a single policy enforcement point may not scale to all agentic interactions.","excerpt":"placing a single entity and expecting it to scale towards all agentic interactions is underestimating the problem space.","support":"source-stated"}],"limitations":["The article is brought to readers by Queen’s University Belfast, but the supplied text does not identify an individual author.","The article describes the de-authentication result as an in-house experiment but does not give its setup, measurement method, or conditions.","The article raises possible approaches, including AI-bill of material and distributed ledger technology, but the supplied text does not report results demonstrating their effectiveness."],"mediaId":"4f2d229e2ac04f8a3d9366c918a446f1f745fee8d9cc033e94a07638ac3f8782","correctionReason":null,"id":"2951cacb0fa934280c51c80980e0f9e08391b1172f52a899c19c4af1e2c43e70","publicationId":"3cde7f25-965a-4456-8fb1-d8fd57729d65","version":1,"sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","sourceTitle":"Can We Trust AI Agents When Zero Trust Isn’t Really Zero","publisher":"IEEE Spectrum","publishedAt":"2026-11-09T16:25:06+00:00","originalPublished":{"value":"2026-11-09T16:25:06+00:00","precision":"instant","basis":"publisher_page"},"firstObservedAt":"2026-11-09T18:27:37.407Z","createdAt":"2026-11-10T00:01:09.505Z","historical":false,"entityIds":["ieee-spectrum"],"topicIds":["power-and-infrastructure","work-and-agency"],"media":[{"id":"4f2d229e2ac04f8a3d9366c918a446f1f745fee8d9cc033e94a07638ac3f8782","kind":"article","url":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-09T21:16:41.384Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."},{"id":"482ccfab586cfaf46252ec9a28049e2a8b31aa2215c65e1edb8823db8dec2597","kind":"image","url":"https://spectrum.ieee.org/media-library/collage-of-indoor-drone-flight-phone-controls-and-operator-at-computer-workstation.jpg?id=68578379&width=1200&height=600&coordinates=0%2C80%2C0%2C80","sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","label":"View source image","provenance":"publisher_page","verification":{"status":"verified","checkedAt":"2026-11-09T21:16:41.384Z","method":"publisher_association_and_head","contentType":"image/jpeg","cacheUntil":1794259301384},"recommendationReason":"Observed on the acquired publisher source; availability checked."}],"sourceMedia":{"status":"found","sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","checkedAt":"2026-11-09T21:16:40.862Z","items":[{"type":"image","url":"https://spectrum.ieee.org/media-library/collage-of-indoor-drone-flight-phone-controls-and-operator-at-computer-workstation.jpg?id=68578379&width=1200&height=600&coordinates=0%2C80%2C0%2C80","sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","attribution":"IEEE Spectrum","basis":"publisher-metadata","availability":"verified","verifiedAt":"2026-11-09T21:16:41.384Z"}],"note":"Media declared by the original publisher; availability checked separately from the editorial reading."},"editorialDecision":{"outcome":"accept","reasonCodes":["useful_perspective"],"rubricVersion":"stream-admission-v5","mediaId":"4f2d229e2ac04f8a3d9366c918a446f1f745fee8d9cc033e94a07638ac3f8782","confidence":0.6900000000000001,"provider":"typesafe","model":"jev-1.13.0","inputHash":"009ff0fd8fda0bd13742e525a034b056e1d33862bc98f08b16e8bcd33c183f30","decidedAt":"2026-11-09T21:16:43.685Z"},"provenance":{"textHash":"e5ed1d230de64fd7897412a5531f2fc6266494d6d7e89d54fca19748cbd30f09","observationId":"d95a4ecccaf2f1f1c050caeb0c686d10abe80aa8df5ed0712ae9dcae8067ce80","retrievedAt":"2026-11-09T21:16:40.862Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"dd2e1783-058d-46de-824c-a13e6e73bd0c"},"reviewer":{"model":"gpt-6-luna","runId":"ca3a1925-e846-47d1-8036-4dbd029a7caf"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-10T00:01:09.505Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Vishal Sharma","url":"https://spectrum.ieee.org/u/vishal-sharma2","kind":"Person"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Vishal Sharma","url":"https://spectrum.ieee.org/u/vishal-sharma2","kind":"Person","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Vishal Sharma","status":"stated","provenance":{"basis":"article-jsonld","sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","locator":"document/html[2]/head[0]/script[0]@163.author"},"assertionId":"bd7efb0b86de563d4ea42bb925973e1228e915af0f68b6748338580d7323d0bf:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T21:16:40.862Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"bd7efb0b86de563d4ea42bb925973e1228e915af0f68b6748338580d7323d0bf","sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","contentSha256":"a5a5a40a7297b60e4100433d07bb64265e57888491777e25bee87218043ad7bd","textHash":"e5ed1d230de64fd7897412a5531f2fc6266494d6d7e89d54fca19748cbd30f09","observedAt":"2026-11-09T21:16:40.862Z"}},{"actor":{"name":"IEEE Spectrum","url":null,"kind":"unknown","actorId":null},"role":"publisher","scope":{"kind":"publication","locator":null},"literalCredit":"IEEE Spectrum","status":"stated","provenance":{"basis":"article-publisher","sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","locator":"document/html[2]/head[0]/script[0]@163.publisher"},"assertionId":"bd7efb0b86de563d4ea42bb925973e1228e915af0f68b6748338580d7323d0bf:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T21:16:40.862Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"bd7efb0b86de563d4ea42bb925973e1228e915af0f68b6748338580d7323d0bf","sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","contentSha256":"a5a5a40a7297b60e4100433d07bb64265e57888491777e25bee87218043ad7bd","textHash":"e5ed1d230de64fd7897412a5531f2fc6266494d6d7e89d54fca19748cbd30f09","observedAt":"2026-11-09T21:16:40.862Z"}}],"participationCoverage":{"total":2,"returned":2,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":{"name":"IEEE Spectrum","url":null},"sourceUrl":"https://spectrum.ieee.org/zero-trust-agentic-ai-architecture","snapshotId":"bd7efb0b86de563d4ea42bb925973e1228e915af0f68b6748338580d7323d0bf","contentSha256":"a5a5a40a7297b60e4100433d07bb64265e57888491777e25bee87218043ad7bd","textHash":"e5ed1d230de64fd7897412a5531f2fc6266494d6d7e89d54fca19748cbd30f09","observedAt":"2026-11-09T21:16:40.862Z","basis":"publisher-page-metadata","sourceLinks":13,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/2951cacb0fa934280c51c80980e0f9e08391b1172f52a899c19c4af1e2c43e70/source"}}}],"audits":[{"publicationId":"3cde7f25-965a-4456-8fb1-d8fd57729d65","verdict":"uncertain","findings":[{"dimension":"attribution","status":"pass","explanation":"The reading correctly identifies this as an IEEE Spectrum article brought to readers by Queen’s University Belfast. It attributes the in-house experiment to QUB and correctly notes that the supplied text names no individual author.","excerpt":"This article is brought to you by Queen’s University Belfast."},{"dimension":"summary","status":"uncertain","explanation":"The summary never explains zero trust ('never trust, always verify'), even though the title and the article's central argument depend on it. Other key terms are also left undefined: de-authentication attack, vibe coding, stateless systems, asynchronous autonomy, horizontal anomalies and AI-bill of material. A general reader cannot easily follow the argument without them. Smallest repair: add the source's own definition of zero trust and tooltips for these terms. The question is fair and follows from the evidence.","excerpt":"all agentic interactions use the core principle of “never trust,” “always verify.”"},{"dimension":"date","status":"pass","explanation":"The originalPublished value, 2026-11-09, comes from the publisher page and matches the feed entry. Nothing in the evidence points to an earlier first publication of this article. The cited references are separate works.","excerpt":null},{"dimension":"support","status":"uncertain","explanation":"Two passages need repair. First, the summary says the experiment 'found agentic orchestration brought' attacks down to 2 to 3 seconds. The source hedges this as an initial understanding that the rate 'could go as low as' 2 to 3 seconds. Fix: restore 'could'. Second, 'These examples support the article's concern' that recovery extends beyond one task or organization is an unlabeled Escape Meridian link. The source states that concern separately and does not tie it to the vibe coding or de-authentication examples. Fix: attribute the concern directly to the article.","excerpt":"provided an initial understanding of how the scale of agentic attacks could mean a lower window of vulnerability"},{"dimension":"media","status":"pass","explanation":"The selected media is the verified direct link to the original IEEE Spectrum article, which is the most relevant way to reach the source.","excerpt":null}],"createdAt":"2026-11-10T00:01:09.505Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}