AI news videos

NVIDIA: Physical AI Needs Safety at Every Layer, and NVIDIA Sells Every Layer

NVIDIA published a new overview of physical AI safety, and the thing worth noticing is that NVIDIA sells every layer it describes.

Watch on YouTube

Transcript

Sources

Every claim in this video comes from the top ranking coverage of this topic. The claims and where each one came from:

• COVERS: HOOK SENTENCE ONE. NARRATED: "NVIDIA says physical AI needs safety at every layer." SUPPORT: article.txt line 67, the headline 'Why Deploying Physical AI at Scale Demands Safety at Every Layer', restated at line 70, 'safety must extend across every layer of the system', and at line 190, 'That requires safety across the hardware, software, AI, operating environment and deployment lifecycle'. WORDING: 'NVIDIA says' is attached from the first word because this is the publisher's own thesis about its own market and is nowhere independently stated on the page; the article's five-item list at line 190 is never narrated, since a comma series speeds the read and breaks the duration budget; 'every layer' is the article's own phrase, not a metaphor supplied by us.(First-party NVIDIA corporate blog post titled 'Why Deploying Physical AI at Scale Demands Safety at Every Layer', written by Riccardo Mariani and dated September 21 2026 (article.txt line 3 for the title, line 67 for the headline, line 74 for the date and line 78 for the byline; the alert-time snapshot at line 2 was captured 2026-09-21T16:04:28.294Z). NVIDIA is both the publisher of this page and the vendor that sells every product described on it, so the page is corporate positioning rather than reporting. The post announces NO new product: it is an overview that consolidates the NVIDIA Halos safety stack NVIDIA already sells, which has been public since at least September 2025, per timeline-probes.json, which records an NVIDIA Omniverse video 'Safety in the Loop: Accelerating Physical AI Certification With NVIDIA Halos' published 2025-09-23 and a third-party mention of 'Halos Robots' dated 2026-06-22; for that reason no launch verb appears anywhere in this script. Only two classes of claim on the page carry outside backing. The market figures are third-party analyst projections: ABI Research projects an installed base of 49 million level 3-5 autonomous vehicles by 2035 (lines 179 to 181) and Omdia estimates roughly 60 million industrial robots deployed between 2026 and 2035 (lines 182 to 184). The assessments are genuine independent third-party work: TUV SUD certified NVIDIA's Automotive Product Lifecycle software process and DriveOS 6.0 to ISO 26262 ASIL D and NVIDIA's automotive engineering processes to ISO/SAE 21434 (lines 306 to 308); TUV Rheinland performed an independent UNECE safety assessment of NVIDIA DRIVE AV and is inspecting IGX Thor, Halos OS and Holoscan Sensor Bridge for functional-safety certification readiness (lines 309 to 314); and ANAB has accredited the NVIDIA Halos AI Systems Inspection Lab as an ISO/IEC 17020 inspection body (lines 317 to 319). Everything else on the page, including the line 212 claim that Halos is 'the first and only full-stack safety system for physical AI', is NVIDIA describing its own products with no outside verification. transcripts.json for this slug is empty (0 videos captured), so there is no transcript corroboration; timeline-probes.json returned mostly unrelated consumer-tech clips and was used only to establish that Halos pre-existed this post, never as a fact source.)
• COVERS: HOOK SENTENCE TWO. NARRATED: "NVIDIA sells every layer." SUPPORT: the Halos stack enumerated at article.txt lines 213 to 254 is entirely NVIDIA product: DRIVE AGX Thor and Hyperion at lines 214 to 218, Halos OS, Halos Core and Halos Middleware at lines 219 to 222, Alpamayo at lines 223 to 226, the Halos Safety Evaluation Framework at lines 227 to 230, IGX Thor at lines 233 to 236, Halos Core for IGX at lines 237 to 238, Holoscan Sensor Bridge at lines 239 to 242, and Isaac Lab plus Omniverse libraries at lines 243 to 249. WORDING: the hook's disclosure beat is placed second so the vendor relationship qualifies the thesis immediately; it is deliberately short and figure-free, and it gives the hookCardWave2 renderer the second beat it requires; 'sells' is the plainest available verb and no product is named, since the glued brand-plus-digit tokens on the page are a text-to-speech risk.(First-party NVIDIA corporate blog post titled 'Why Deploying Physical AI at Scale Demands Safety at Every Layer', written by Riccardo Mariani and dated September 21 2026 (article.txt line 3 for the title, line 67 for the headline, line 74 for the date and line 78 for the byline; the alert-time snapshot at line 2 was captured 2026-09-21T16:04:28.294Z). NVIDIA is both the publisher of this page and the vendor that sells every product described on it, so the page is corporate positioning rather than reporting. The post announces NO new product: it is an overview that consolidates the NVIDIA Halos safety stack NVIDIA already sells, which has been public since at least September 2025, per timeline-probes.json, which records an NVIDIA Omniverse video 'Safety in the Loop: Accelerating Physical AI Certification With NVIDIA Halos' published 2025-09-23 and a third-party mention of 'Halos Robots' dated 2026-06-22; for that reason no launch verb appears anywhere in this script. Only two classes of claim on the page carry outside backing. The market figures are third-party analyst projections: ABI Research projects an installed base of 49 million level 3-5 autonomous vehicles by 2035 (lines 179 to 181) and Omdia estimates roughly 60 million industrial robots deployed between 2026 and 2035 (lines 182 to 184). The assessments are genuine independent third-party work: TUV SUD certified NVIDIA's Automotive Product Lifecycle software process and DriveOS 6.0 to ISO 26262 ASIL D and NVIDIA's automotive engineering processes to ISO/SAE 21434 (lines 306 to 308); TUV Rheinland performed an independent UNECE safety assessment of NVIDIA DRIVE AV and is inspecting IGX Thor, Halos OS and Holoscan Sensor Bridge for functional-safety certification readiness (lines 309 to 314); and ANAB has accredited the NVIDIA Halos AI Systems Inspection Lab as an ISO/IEC 17020 inspection body (lines 317 to 319). Everything else on the page, including the line 212 claim that Halos is 'the first and only full-stack safety system for physical AI', is NVIDIA describing its own products with no outside verification. transcripts.json for this slug is empty (0 videos captured), so there is no transcript corroboration; timeline-probes.json returned mostly unrelated consumer-tech clips and was used only to establish that Halos pre-existed this post, never as a fact source.)
• COVERS: SCENE 1 SENTENCE ONE. NARRATED: "ABI Research projects forty nine million autonomous vehicles by twenty thirty five." SUPPORT: article.txt lines 179 to 181, 'By 2035, ABI Research projects an installed base of 49 million level 3-5 autonomous vehicles (AVs)'. WORDING: the forecast is attributed to ABI Research and never to NVIDIA, because it is the only market figure on the page with an outside author; 'installed base of level 3-5' is compressed to 'autonomous vehicles' because 'level 3-5' is a hyphenated brand-plus-digit token that the text-to-speech garbles and the automation-level distinction is not load-bearing for the point being made; the numbers are spoken as words per house style and appear as digits only on the title card; the neighbouring Omdia figure of roughly 60 million industrial robots at lines 182 to 184 is deliberately NOT narrated, because two large figures spoken close together send the speech-to-text auditor into a repetition loop and fail the render.(First-party NVIDIA corporate blog post titled 'Why Deploying Physical AI at Scale Demands Safety at Every Layer', written by Riccardo Mariani and dated September 21 2026 (article.txt line 3 for the title, line 67 for the headline, line 74 for the date and line 78 for the byline; the alert-time snapshot at line 2 was captured 2026-09-21T16:04:28.294Z). NVIDIA is both the publisher of this page and the vendor that sells every product described on it, so the page is corporate positioning rather than reporting. The post announces NO new product: it is an overview that consolidates the NVIDIA Halos safety stack NVIDIA already sells, which has been public since at least September 2025, per timeline-probes.json, which records an NVIDIA Omniverse video 'Safety in the Loop: Accelerating Physical AI Certification With NVIDIA Halos' published 2025-09-23 and a third-party mention of 'Halos Robots' dated 2026-06-22; for that reason no launch verb appears anywhere in this script. Only two classes of claim on the page carry outside backing. The market figures are third-party analyst projections: ABI Research projects an installed base of 49 million level 3-5 autonomous vehicles by 2035 (lines 179 to 181) and Omdia estimates roughly 60 million industrial robots deployed between 2026 and 2035 (lines 182 to 184). The assessments are genuine independent third-party work: TUV SUD certified NVIDIA's Automotive Product Lifecycle software process and DriveOS 6.0 to ISO 26262 ASIL D and NVIDIA's automotive engineering processes to ISO/SAE 21434 (lines 306 to 308); TUV Rheinland performed an independent UNECE safety assessment of NVIDIA DRIVE AV and is inspecting IGX Thor, Halos OS and Holoscan Sensor Bridge for functional-safety certification readiness (lines 309 to 314); and ANAB has accredited the NVIDIA Halos AI Systems Inspection Lab as an ISO/IEC 17020 inspection body (lines 317 to 319). Everything else on the page, including the line 212 claim that Halos is 'the first and only full-stack safety system for physical AI', is NVIDIA describing its own products with no outside verification. transcripts.json for this slug is empty (0 videos captured), so there is no transcript corroboration; timeline-probes.json returned mostly unrelated consumer-tech clips and was used only to establish that Halos pre-existed this post, never as a fact source.)
• COVERS: SCENE 1 SENTENCE TWO. NARRATED: "They will operate in spaces shared with people." SUPPORT: article.txt line 185, 'As these machines enter roads, factories, warehouses and other environments shared with people, safety must scale with them', and line 196, 'autonomous machines move into factories, warehouses and other environments shared with people'. WORDING: the article's three-item list of environments is collapsed to the single word 'spaces' so no comma series appears in narration; 'operate in' replaces 'enter', which the article uses transitively across a list; the sentence carries no figure, which separates the forty nine million from the next scene.(First-party NVIDIA corporate blog post titled 'Why Deploying Physical AI at Scale Demands Safety at Every Layer', written by Riccardo Mariani and dated September 21 2026 (article.txt line 3 for the title, line 67 for the headline, line 74 for the date and line 78 for the byline; the alert-time snapshot at line 2 was captured 2026-09-21T16:04:28.294Z). NVIDIA is both the publisher of this page and the vendor that sells every product described on it, so the page is corporate positioning rather than reporting. The post announces NO new product: it is an overview that consolidates the NVIDIA Halos safety stack NVIDIA already sells, which has been public since at least September 2025, per timeline-probes.json, which records an NVIDIA Omniverse video 'Safety in the Loop: Accelerating Physical AI Certification With NVIDIA Halos' published 2025-09-23 and a third-party mention of 'Halos Robots' dated 2026-06-22; for that reason no launch verb appears anywhere in this script. Only two classes of claim on the page carry outside backing. The market figures are third-party analyst projections: ABI Research projects an installed base of 49 million level 3-5 autonomous vehicles by 2035 (lines 179 to 181) and Omdia estimates roughly 60 million industrial robots deployed between 2026 and 2035 (lines 182 to 184). The assessments are genuine independent third-party work: TUV SUD certified NVIDIA's Automotive Product Lifecycle software process and DriveOS 6.0 to ISO 26262 ASIL D and NVIDIA's automotive engineering processes to ISO/SAE 21434 (lines 306 to 308); TUV Rheinland performed an independent UNECE safety assessment of NVIDIA DRIVE AV and is inspecting IGX Thor, Halos OS and Holoscan Sensor Bridge for functional-safety certification readiness (lines 309 to 314); and ANAB has accredited the NVIDIA Halos AI Systems Inspection Lab as an ISO/IEC 17020 inspection body (lines 317 to 319). Everything else on the page, including the line 212 claim that Halos is 'the first and only full-stack safety system for physical AI', is NVIDIA describing its own products with no outside verification. transcripts.json for this slug is empty (0 videos captured), so there is no transcript corroboration; timeline-probes.json returned mostly unrelated consumer-tech clips and was used only to establish that Halos pre-existed this post, never as a fact source.)
• COVERS: SCENE 2 SENTENCE ONE. NARRATED: "A single check before deployment no longer works." SUPPORT: article.txt line 190, 'That requires safety across the hardware, software, AI, operating environment and deployment lifecycle - not a one-time check before deployment', reinforced at line 208, 'these shifts require safety to be operationalized across design, deployment and validation'. WORDING: the article's negative construction is turned into a plain positive sentence; 'one-time check' becomes 'a single check' to avoid a hyphenated compound, which the text-to-speech stutters through; the sentence names nothing and lists nothing.(First-party NVIDIA corporate blog post titled 'Why Deploying Physical AI at Scale Demands Safety at Every Layer', written by Riccardo Mariani and dated September 21 2026 (article.txt line 3 for the title, line 67 for the headline, line 74 for the date and line 78 for the byline; the alert-time snapshot at line 2 was captured 2026-09-21T16:04:28.294Z). NVIDIA is both the publisher of this page and the vendor that sells every product described on it, so the page is corporate positioning rather than reporting. The post announces NO new product: it is an overview that consolidates the NVIDIA Halos safety stack NVIDIA already sells, which has been public since at least September 2025, per timeline-probes.json, which records an NVIDIA Omniverse video 'Safety in the Loop: Accelerating Physical AI Certification With NVIDIA Halos' published 2025-09-23 and a third-party mention of 'Halos Robots' dated 2026-06-22; for that reason no launch verb appears anywhere in this script. Only two classes of claim on the page carry outside backing. The market figures are third-party analyst projections: ABI Research projects an installed base of 49 million level 3-5 autonomous vehicles by 2035 (lines 179 to 181) and Omdia estimates roughly 60 million industrial robots deployed between 2026 and 2035 (lines 182 to 184). The assessments are genuine independent third-party work: TUV SUD certified NVIDIA's Automotive Product Lifecycle software process and DriveOS 6.0 to ISO 26262 ASIL D and NVIDIA's automotive engineering processes to ISO/SAE 21434 (lines 306 to 308); TUV Rheinland performed an independent UNECE safety assessment of NVIDIA DRIVE AV and is inspecting IGX Thor, Halos OS and Holoscan Sensor Bridge for functional-safety certification readiness (lines 309 to 314); and ANAB has accredited the NVIDIA Halos AI Systems Inspection Lab as an ISO/IEC 17020 inspection body (lines 317 to 319). Everything else on the page, including the line 212 claim that Halos is 'the first and only full-stack safety system for physical AI', is NVIDIA describing its own products with no outside verification. transcripts.json for this slug is empty (0 videos captured), so there is no transcript corroboration; timeline-probes.json returned mostly unrelated consumer-tech clips and was used only to establish that Halos pre-existed this post, never as a fact source.)
• COVERS: SCENE 2 SENTENCE TWO. NARRATED: "Software keeps updating and environments keep changing afterward." SUPPORT: article.txt lines 204 to 205, 'Deployment is ongoing. AVs and robots evolve through software and model updates, new tasks and changing operating conditions. Material changes may require additional safety testing', and lines 200 to 201, 'Dynamic environments require context-aware safety. Roads, factories and warehouses cannot be fully controlled through static zones or physical barriers'. WORDING: two causes are joined by 'and', never by a comma, which is the maximum this house style allows in one sentence; the article's own lists of update types and environment types are dropped entirely; 'afterward' carries the article's 'ongoing' and 'after deployment' sense in one word without repeating 'deployment' from the previous sentence.(First-party NVIDIA corporate blog post titled 'Why Deploying Physical AI at Scale Demands Safety at Every Layer', written by Riccardo Mariani and dated September 21 2026 (article.txt line 3 for the title, line 67 for the headline, line 74 for the date and line 78 for the byline; the alert-time snapshot at line 2 was captured 2026-09-21T16:04:28.294Z). NVIDIA is both the publisher of this page and the vendor that sells every product described on it, so the page is corporate positioning rather than reporting. The post announces NO new product: it is an overview that consolidates the NVIDIA Halos safety stack NVIDIA already sells, which has been public since at least September 2025, per timeline-probes.json, which records an NVIDIA Omniverse video 'Safety in the Loop: Accelerating Physical AI Certification With NVIDIA Halos' published 2025-09-23 and a third-party mention of 'Halos Robots' dated 2026-06-22; for that reason no launch verb appears anywhere in this script. Only two classes of claim on the page carry outside backing. The market figures are third-party analyst projections: ABI Research projects an installed base of 49 million level 3-5 autonomous vehicles by 2035 (lines 179 to 181) and Omdia estimates roughly 60 million industrial robots deployed between 2026 and 2035 (lines 182 to 184). The assessments are genuine independent third-party work: TUV SUD certified NVIDIA's Automotive Product Lifecycle software process and DriveOS 6.0 to ISO 26262 ASIL D and NVIDIA's automotive engineering processes to ISO/SAE 21434 (lines 306 to 308); TUV Rheinland performed an independent UNECE safety assessment of NVIDIA DRIVE AV and is inspecting IGX Thor, Halos OS and Holoscan Sensor Bridge for functional-safety certification readiness (lines 309 to 314); and ANAB has accredited the NVIDIA Halos AI Systems Inspection Lab as an ISO/IEC 17020 inspection body (lines 317 to 319). Everything else on the page, including the line 212 claim that Halos is 'the first and only full-stack safety system for physical AI', is NVIDIA describing its own products with no outside verification. transcripts.json for this slug is empty (0 videos captured), so there is no transcript corroboration; timeline-probes.json returned mostly unrelated consumer-tech clips and was used only to establish that Halos pre-existed this post, never as a fact source.)
• COVERS: SCENE 3 SENTENCE ONE. NARRATED: "Halos is NVIDIA's full stack safety system covering the chip and the simulation that validates behavior." SUPPORT: article.txt lines 211 to 212, 'NVIDIA Halos is the first and only full-stack safety system for physical AI, helping developers engineer safety across every layer of design, validation and deployment'; the chip layer at lines 233 to 236, 'NVIDIA IGX Thor is an industrial-grade module that combines accelerated computing and functional safety on one platform'; and the simulation layer at lines 243 to 249, 'NVIDIA Isaac Lab and NVIDIA Omniverse libraries let developers test robot behavior across relevant conditions and edge cases'. WORDING: NVIDIA's 'first and only' superlative is NOT narrated, because it is an unverified vendor claim and repeating it would launder marketing copy into reporting; 'full stack' is written without a hyphen for the text-to-speech; the stack is reduced to exactly two endpoints joined by 'and', instead of the article's four-part component list, and the operating-system layer at lines 219 to 222 is dropped to stay inside the word cap; no product name containing digits is spoken.(First-party NVIDIA corporate blog post titled 'Why Deploying Physical AI at Scale Demands Safety at Every Layer', written by Riccardo Mariani and dated September 21 2026 (article.txt line 3 for the title, line 67 for the headline, line 74 for the date and line 78 for the byline; the alert-time snapshot at line 2 was captured 2026-09-21T16:04:28.294Z). NVIDIA is both the publisher of this page and the vendor that sells every product described on it, so the page is corporate positioning rather than reporting. The post announces NO new product: it is an overview that consolidates the NVIDIA Halos safety stack NVIDIA already sells, which has been public since at least September 2025, per timeline-probes.json, which records an NVIDIA Omniverse video 'Safety in the Loop: Accelerating Physical AI Certification With NVIDIA Halos' published 2025-09-23 and a third-party mention of 'Halos Robots' dated 2026-06-22; for that reason no launch verb appears anywhere in this script. Only two classes of claim on the page carry outside backing. The market figures are third-party analyst projections: ABI Research projects an installed base of 49 million level 3-5 autonomous vehicles by 2035 (lines 179 to 181) and Omdia estimates roughly 60 million industrial robots deployed between 2026 and 2035 (lines 182 to 184). The assessments are genuine independent third-party work: TUV SUD certified NVIDIA's Automotive Product Lifecycle software process and DriveOS 6.0 to ISO 26262 ASIL D and NVIDIA's automotive engineering processes to ISO/SAE 21434 (lines 306 to 308); TUV Rheinland performed an independent UNECE safety assessment of NVIDIA DRIVE AV and is inspecting IGX Thor, Halos OS and Holoscan Sensor Bridge for functional-safety certification readiness (lines 309 to 314); and ANAB has accredited the NVIDIA Halos AI Systems Inspection Lab as an ISO/IEC 17020 inspection body (lines 317 to 319). Everything else on the page, including the line 212 claim that Halos is 'the first and only full-stack safety system for physical AI', is NVIDIA describing its own products with no outside verification. transcripts.json for this slug is empty (0 videos captured), so there is no transcript corroboration; timeline-probes.json returned mostly unrelated consumer-tech clips and was used only to establish that Halos pre-existed this post, never as a fact source.)
• COVERS: SCENE 3 SENTENCE TWO. NARRATED: "Agility is building it into its humanoid robot." SUPPORT: article.txt lines 301 to 304, 'Agility is integrating NVIDIA IGX Thor and Halos Core into the safety system for its Digit 5 humanoid'. WORDING: exactly one customer example is named, as instructed; the KION Group autonomous-forklift work at lines 299 to 300 and the long partner rosters at lines 261 to 298 are all omitted, because those are comma series and because a roster of logos is the least verifiable part of a vendor page; 'Digit 5' is not spoken, since the glued brand-plus-digit token is a known text-to-speech garble; 'building it into' compresses 'integrating into the safety system for', and 'it' refers to the Halos stack named in the preceding sentence.(First-party NVIDIA corporate blog post titled 'Why Deploying Physical AI at Scale Demands Safety at Every Layer', written by Riccardo Mariani and dated September 21 2026 (article.txt line 3 for the title, line 67 for the headline, line 74 for the date and line 78 for the byline; the alert-time snapshot at line 2 was captured 2026-09-21T16:04:28.294Z). NVIDIA is both the publisher of this page and the vendor that sells every product described on it, so the page is corporate positioning rather than reporting. The post announces NO new product: it is an overview that consolidates the NVIDIA Halos safety stack NVIDIA already sells, which has been public since at least September 2025, per timeline-probes.json, which records an NVIDIA Omniverse video 'Safety in the Loop: Accelerating Physical AI Certification With NVIDIA Halos' published 2025-09-23 and a third-party mention of 'Halos Robots' dated 2026-06-22; for that reason no launch verb appears anywhere in this script. Only two classes of claim on the page carry outside backing. The market figures are third-party analyst projections: ABI Research projects an installed base of 49 million level 3-5 autonomous vehicles by 2035 (lines 179 to 181) and Omdia estimates roughly 60 million industrial robots deployed between 2026 and 2035 (lines 182 to 184). The assessments are genuine independent third-party work: TUV SUD certified NVIDIA's Automotive Product Lifecycle software process and DriveOS 6.0 to ISO 26262 ASIL D and NVIDIA's automotive engineering processes to ISO/SAE 21434 (lines 306 to 308); TUV Rheinland performed an independent UNECE safety assessment of NVIDIA DRIVE AV and is inspecting IGX Thor, Halos OS and Holoscan Sensor Bridge for functional-safety certification readiness (lines 309 to 314); and ANAB has accredited the NVIDIA Halos AI Systems Inspection Lab as an ISO/IEC 17020 inspection body (lines 317 to 319). Everything else on the page, including the line 212 claim that Halos is 'the first and only full-stack safety system for physical AI', is NVIDIA describing its own products with no outside verification. transcripts.json for this slug is empty (0 videos captured), so there is no transcript corroboration; timeline-probes.json returned mostly unrelated consumer-tech clips and was used only to establish that Halos pre-existed this post, never as a fact source.)
• COVERS: SCENE 4 SENTENCE ONE. NARRATED: "NVIDIA wrote this overview of products it already sells and announced nothing new." SUPPORT: the publisher is established at article.txt line 1, 'SOURCE URL: https://blogs.nvidia.com/blog/physical-ai-halos-safety/', with the byline at line 78 and the date at line 74; the page contains no launch verb, no release date and no availability statement anywhere, and closes at lines 321 to 325 with a marketing call to action, 'Learn more about NVIDIA Halos for AVs and robotics'; the products described are pre-existing, per timeline-probes.json, which dates a public NVIDIA Halos certification video to 2025-09-23 and a third-party 'Halos Robots' mention to 2026-06-22. WORDING: this is the mandatory honesty beat and it names the publisher outright rather than using a pronoun; 'wrote this overview' and 'announced nothing new' are used precisely because the alert tier would otherwise invite launch language, and no form of launched, unveiled, announced today, new or just released is applied to Halos or any component anywhere in this script.(First-party NVIDIA corporate blog post titled 'Why Deploying Physical AI at Scale Demands Safety at Every Layer', written by Riccardo Mariani and dated September 21 2026 (article.txt line 3 for the title, line 67 for the headline, line 74 for the date and line 78 for the byline; the alert-time snapshot at line 2 was captured 2026-09-21T16:04:28.294Z). NVIDIA is both the publisher of this page and the vendor that sells every product described on it, so the page is corporate positioning rather than reporting. The post announces NO new product: it is an overview that consolidates the NVIDIA Halos safety stack NVIDIA already sells, which has been public since at least September 2025, per timeline-probes.json, which records an NVIDIA Omniverse video 'Safety in the Loop: Accelerating Physical AI Certification With NVIDIA Halos' published 2025-09-23 and a third-party mention of 'Halos Robots' dated 2026-06-22; for that reason no launch verb appears anywhere in this script. Only two classes of claim on the page carry outside backing. The market figures are third-party analyst projections: ABI Research projects an installed base of 49 million level 3-5 autonomous vehicles by 2035 (lines 179 to 181) and Omdia estimates roughly 60 million industrial robots deployed between 2026 and 2035 (lines 182 to 184). The assessments are genuine independent third-party work: TUV SUD certified NVIDIA's Automotive Product Lifecycle software process and DriveOS 6.0 to ISO 26262 ASIL D and NVIDIA's automotive engineering processes to ISO/SAE 21434 (lines 306 to 308); TUV Rheinland performed an independent UNECE safety assessment of NVIDIA DRIVE AV and is inspecting IGX Thor, Halos OS and Holoscan Sensor Bridge for functional-safety certification readiness (lines 309 to 314); and ANAB has accredited the NVIDIA Halos AI Systems Inspection Lab as an ISO/IEC 17020 inspection body (lines 317 to 319). Everything else on the page, including the line 212 claim that Halos is 'the first and only full-stack safety system for physical AI', is NVIDIA describing its own products with no outside verification. transcripts.json for this slug is empty (0 videos captured), so there is no transcript corroboration; timeline-probes.json returned mostly unrelated consumer-tech clips and was used only to establish that Halos pre-existed this post, never as a fact source.)
• COVERS: SCENE 4 SENTENCE TWO. NARRATED: "Only the analyst forecast and the independent certifications come from outside NVIDIA." SUPPORT: the outside-backed items are the ABI Research projection at lines 179 to 181 and the Omdia estimate at lines 182 to 184, both third-party analyst firms; and the independent assessments at lines 305 to 319, namely TUV SUD's ISO 26262 ASIL D certification of NVIDIA's software process and DriveOS 6.0 plus its ISO/SAE 21434 certification of NVIDIA's automotive engineering processes, TUV Rheinland's independent UNECE safety assessment of NVIDIA DRIVE AV and its in-progress functional-safety inspection of IGX Thor, Halos OS and Holoscan Sensor Bridge, and ANAB's accreditation of the NVIDIA Halos AI Systems Inspection Lab as an ISO/IEC 17020 inspection body. Every other claim on the page is NVIDIA describing NVIDIA products. WORDING: 'Only' does the honesty work by bounding what is verified rather than asserting that anything is false, which is the defensible framing; the certifying bodies are not named, because the initialisms and the standard numbers are text-to-speech hazards and naming three of them would be a comma series; 'forecast' is singular because only one forecast is narrated, while 'certifications' is plural because several are genuine; note that TUV Rheinland's robotics inspection is described on the page as in progress rather than complete, which is why the sentence says the certifications come from outside NVIDIA rather than claiming Halos is certified.(First-party NVIDIA corporate blog post titled 'Why Deploying Physical AI at Scale Demands Safety at Every Layer', written by Riccardo Mariani and dated September 21 2026 (article.txt line 3 for the title, line 67 for the headline, line 74 for the date and line 78 for the byline; the alert-time snapshot at line 2 was captured 2026-09-21T16:04:28.294Z). NVIDIA is both the publisher of this page and the vendor that sells every product described on it, so the page is corporate positioning rather than reporting. The post announces NO new product: it is an overview that consolidates the NVIDIA Halos safety stack NVIDIA already sells, which has been public since at least September 2025, per timeline-probes.json, which records an NVIDIA Omniverse video 'Safety in the Loop: Accelerating Physical AI Certification With NVIDIA Halos' published 2025-09-23 and a third-party mention of 'Halos Robots' dated 2026-06-22; for that reason no launch verb appears anywhere in this script. Only two classes of claim on the page carry outside backing. The market figures are third-party analyst projections: ABI Research projects an installed base of 49 million level 3-5 autonomous vehicles by 2035 (lines 179 to 181) and Omdia estimates roughly 60 million industrial robots deployed between 2026 and 2035 (lines 182 to 184). The assessments are genuine independent third-party work: TUV SUD certified NVIDIA's Automotive Product Lifecycle software process and DriveOS 6.0 to ISO 26262 ASIL D and NVIDIA's automotive engineering processes to ISO/SAE 21434 (lines 306 to 308); TUV Rheinland performed an independent UNECE safety assessment of NVIDIA DRIVE AV and is inspecting IGX Thor, Halos OS and Holoscan Sensor Bridge for functional-safety certification readiness (lines 309 to 314); and ANAB has accredited the NVIDIA Halos AI Systems Inspection Lab as an ISO/IEC 17020 inspection body (lines 317 to 319). Everything else on the page, including the line 212 claim that Halos is 'the first and only full-stack safety system for physical AI', is NVIDIA describing its own products with no outside verification. transcripts.json for this slug is empty (0 videos captured), so there is no transcript corroboration; timeline-probes.json returned mostly unrelated consumer-tech clips and was used only to establish that Halos pre-existed this post, never as a fact source.)