View source for User:Louis/California AB 2047 and the Physna conflict of interest

{{ombox|text=This is a personal, signed analysis by Louis Rossmann, kept in his userspace. It argues a position and is not a neutral mainspace article.}}

California Assembly Bill 2047, the California Firearm Printing Prevention Act, would require every 3D printer sold in the state to run a state-mandated detection algorithm that reads each print file and blocks anything it decides would produce a firearm or an illegal firearm part.<ref name="bill" /> The bill passed the Assembly on May 26, 2026 by a vote of 58 to 19 and moved to the Senate the next day.<ref name="bill-history" /><ref name="toms-ca" /> One company has gone on the public record telling lawmakers this kind of detection is ready to deploy: Physna, Inc., the Columbus, Ohio firm that built what it calls ''the world's first true 3D geometric search engine.''<ref name="ap-lieb" /><ref name="sbir-geom" /> It is also the company that holds the patents on the matching method a mandate like this would force every printer maker to buy, and its best evidence that the matching works is a set of United States Air Force contracts, where the work runs on enterprise and cloud hardware, not on the controller board inside a desktop printer.<ref name="pat736" /><ref name="pat531" /><ref name="usaf-geom" /> I oppose this bill, on the same technical, moral, and philosophical grounds I oppose the [[New York 3D printer blocking technology mandate]].

== What AB 2047 would require ==

AB 2047, as amended in the Senate on June 15, 2026, adds Title 21.1 to the Civil Code and a new section to the Penal Code.<ref name="bill" /> It defines a ''firearm blueprint detection algorithm'' as a software service that evaluates a printing file, in the form of an STL mesh, a CAD file, or ''geometric code,'' and flags any file it decides would produce a firearm or illegal firearm part.<ref name="bill" /> The California Department of Justice would publish written guidance on performance standards for those detection algorithms on or before September 1, 2028.<ref name="bill" /> The statute even names the technique it is reaching for: it directs the Department to review emerging detection methods ''including, but not limited to ... volumetric search functionality.''<ref name="bill" />

<gallery mode="packed" heights="200" caption="AB 2047, the detection algorithm it defines and the technique it names">
File:CRW-ab2047-leginfo-algorithm-def.png|AB 2047 defines a firearm blueprint detection algorithm as a software service that evaluates a printing file, whether an STL mesh, a CAD file, or geometric code. [https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202520260AB2047 California Legislative Information, AB 2047].
File:CRW-ab2047-leginfo-volumetric.png|AB 2047 directs the state to periodically review emerging detection techniques, naming volumetric search functionality among them. [https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202520260AB2047 California Legislative Information, AB 2047].
</gallery>

The compliance regime lands on the manufacturers. By March 1, 2029 any business that makes printers for sale in California has to file a sworn attestation for each model, and a false attestation is punishable as perjury.<ref name="bill" /> By June 1, 2029 the Department publishes a list of attested models, updated quarterly.<ref name="bill" /> Starting December 1, 2029 it is unlawful to sell a printer that is not equipped and not on the list, with a civil penalty of up to twenty-five thousand dollars per violation.<ref name="bill" /> The bill also adds Penal Code section 29187, which makes it a misdemeanor to knowingly disable or circumvent the blocking technology ''with intent to manufacture firearms,'' and to sell modified versions of a listed printer ''with the intent to facilitate the unlawful manufacture of firearms.''<ref name="bill" /> The Electronic Frontier Foundation argues that clause reaches ordinary owners, because it would ''criminalize the use of open-source alternatives'' like the Marlin and Klipper firmware that most consumer printers run.<ref name="eff-ca" /><ref name="toms-ca" />

== Physna's on-record feasibility claim ==

At an online seminar on this legislation, a Physna technical account manager named Julian Chultarsky told the audience the technology is ready. His exact words, as reported by the Associated Press, were ''Geometric search is mature, it's deployed, it is ready to be applied to this problem.''<ref name="ap-lieb" /> Solomon Diamond, an associate engineering professor at Dartmouth who was on the same panel, described the idea as a bit like a phone app that identifies a tree or a flower from a photo.<ref name="ap-lieb" />

[[File:CRW-ab2047-ap-chultarsky-quote.png|thumb|center|upright=2.4|Physna technical account manager Julian Chultarsky told the Associated Press that geometric search is mature, deployed, and ready to be applied to this problem. [https://www.pbs.org/newshour/nation/first-of-its-kind-law-in-new-york-could-block-3d-printers-from-making-guns Associated Press, via PBS NewsHour].]]

In its 2023 Air Force contract abstract, Physna calls its platform ''the world's first true 3D geometric search engine and deep learning platform,'' built to ''search by geometry'' and label and retrieve 3D models by shape.<ref name="sbir-geom" /> Geometric search of 3D files is exactly the category AB 2047 would have the state set the standard for and then mandate.

[[File:CRW-ab2047-sbir-abstract-engine.png|thumb|center|upright=2.4|In its 2023 Air Force award abstract, Physna describes its platform as the world's first true 3D geometric search engine and deep learning platform. [https://www.sbir.gov/awards/205940 SBIR.gov].]]

== Patents Physna holds on the matching method ==

Physna holds the foundational patents on the method. United States Patent 11,422,531, ''System and methods for 3D model evaluation,'' was granted on August 23, 2022, and United States Patent 12,147,736, ''System and methods for 3D model evaluation using triangle mesh hashing,'' was granted on November 19, 2024.<ref name="pat531" /><ref name="pat736" /> Both are assigned to Physna Inc., and both name Paul Powers, Physna's founder,<ref name="powers-founder" /> and Glenn Warner as the inventors.<ref name="pat531" /><ref name="pat736" /> The 2024 patent describes hashing a model's triangle mesh into a fingerprint and comparing it against reference models, including detecting added or removed sub-parts.<ref name="pat736" />

[[File:CRW-ab2047-patent-12147736.png|thumb|center|upright=2.4|United States Patent 12,147,736 is assigned to Physna Inc. and names Glenn Warner and Paul Powers as inventors. [https://patents.google.com/patent/US12147736B2/en US Patent 12,147,736, via Google Patents].]]

A bill that forces every printer maker to ship a geometric detection algorithm builds a captive market for exactly this patented method. The incentive is documented and structural: the state would require a category of technology, and one company holds the patents and the on-record sales pitch for that category. That is a conflict of interest whether or not a single email ever passed between Physna and the bill's author.

== Why the Air Force proof does not run on a desktop printer ==

Physna's strongest evidence that its matching works is its federal record, and that record is a defense and enterprise record. The Air Force awarded Physna a Small Business Innovation Research Phase II contract, FA8649-23-P-0450, ''Geometric Deep Learning Technology for Supply Chain,'' for $1,040,000 in 2023.<ref name="usaf-geom" /><ref name="sbir-geom" /> It followed with FA7014-25-C-0023, ''Physna SBIR III Technology of Supply,'' for $1,500,000, and a $1,900,000 award for multimodal image search supporting Air Force Digital Material Management.<ref name="usaf-sbir3" /><ref name="usaf-multimodal" /> A separate Air Force order, FA7014-24-P-0043, paid Physna to ''identify alternate sources of supply through comparison of digital files.''<ref name="usaf-altsource" /> The company also carries a General Services Administration schedule contract, 47QTCA24D00BN, for federal information-technology buyers.<ref name="gsa" /> This is geometric matching run on cloud and enterprise hardware, against catalogs of aerospace parts.

A consumer printer is not that. The board that drives a $200 printer's motors does not have the memory or the processor to reconstruct a model from a stream of G-code and hash it against a live national database of firearm files before every job. So the mandate forces one of two outcomes. Either the print file gets uploaded to a server for the check, which turns every printer in the state into a reporting device and puts home fabrication under surveillance, a harm the EFF says the bill would create.<ref name="eff-ca" /> Or the printer carries enough added hardware to do the work locally, which raises the price of the cheapest machines, the ones in schools, libraries, and makerspaces. AB 2047 exempts printers sold to licensed firearms makers, to law enforcement, and to aerospace, automotive, biomedical, and propmaking buyers, but it writes no exemption for the consumer machine on a classroom bench.<ref name="bill" /><ref name="toms-ca" /> Tom's Hardware noted the same gap: detection capable enough to catch a gun part ''may exceed the compute available on consumer printers, pushing checks to remote servers.''<ref name="toms-ca" />

== Shared geometry and the industry's own verdict ==

The detection cannot be made reliable, because a firearm part is a geometric signature shared with ordinary objects. A suppressor baffle is the same stack of cones as a Tesla fluid valve; a pistol conversion device is the same dovetailed block as a camera flash shoe; an AR-15 lower is the same hollowed box as an electronics enclosure; a suppressor tube is the same threaded cylinder as a flashlight body. I documented these geometric twins, file format by file format, in a [[User:Louis/3D printer firearm-blocking mandates and geometric false positives|companion essay]], and the chokepoint problem behind them in [[User:Louis/3D-printed firearms and the technical basis for printer mandates|another]]. Set the filter wide enough to catch the gun part and it flags pipes and brackets; set it narrow and a sub-visible edit or a split file walks straight through.

Adafruit's Phillip Torrone laid out the same problem in Techdirt:
<blockquote>''A firearms blueprint detection algorithm would need to identify every possible firearm component from raw STL/GCODE files, while not flagging pipes, tubes, blocks, brackets, gears, or any of the millions of legitimate shapes that happen to share geometric properties with gun parts. This is a classification problem with enormous false positive and false negative rates.''</blockquote><ref name="techdirt-ca" />
The EFF called the requirement ''censorware'' that is ''doomed to fail for its intended purpose.''<ref name="eff-ca" />

[[File:CRW-ab2047-eff-censorware.png|thumb|center|upright=2.4|The Electronic Frontier Foundation describes AB 2047 as a censorware mandate that would criminalize the use of open-source alternatives. [https://www.eff.org/deeplinks/2026/04/dangers-californias-legislation-censor-3d-printing Electronic Frontier Foundation].]]

AB 2047 concedes the point in its own text: it tells the Department to set ''acceptable false-positive and false-negative rates'' rather than require a perfect one, and it asks stakeholders to identify benign files an algorithm wrongly flags.<ref name="bill" /> Bill Decker, executive chairman of the Association of 3D Printing, whose organization supports the New York and California laws, says of the technology: ''It's not going to work. It's more of a political statement than anything else.''<ref name="ap-lieb" />

== No lobbying paper trail to Physna ==

I went looking for a paper trail tying Physna or Paul Powers to this bill, and there is not one. There is no public record that Physna lobbied for AB 2047, drafted it, paid for it, or was named as its vendor, and none for the New York mandate either. The documented legislative push comes from [[User:Louis/Everytown for Gun Safety and the 3D printer blocking mandates|Everytown for Gun Safety]] and from the bill's author, Assembly Member Rebecca Bauer-Kahan.<ref name="toms-ca" />

None of this requires anyone to have done anything corrupt. A documented conflict of interest and a detection method that has never been shown to run on a consumer printer sit in the same bill, and either one alone is reason enough to vote it down.

== See also ==
* [[New York 3D printer blocking technology mandate]]
* [[3D Printing restrictions and bans]]
* [[User:Louis/3D printer firearm-blocking mandates and geometric false positives]]
* [[User:Louis/3D-printed firearms and the technical basis for printer mandates]]
* [[User:Louis/Everytown for Gun Safety and the 3D printer blocking mandates]]
* [[User:Louis/Bloomberg and the control of 3D printing]]
* [[Right to Repair]]
* [[Digital rights management]]

== References ==
<references>
<ref name="bill">{{Cite web |url=https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202520260AB2047 |title=AB-2047 Firearms: 3-dimensional printing blocking technology, as amended in the Senate June 15, 2026 |publisher=California Legislative Information |date=2026-06-15 |access-date=2026-06-23}} Civil Code Title 21.1 (Section 3273.631 et seq.), the California Firearm Printing Prevention Act; the operative text is headed "Amended IN Senate June 15, 2026." Section 3273.632 defines the "firearm blueprint detection algorithm" evaluating a printing file "whether in the form of stereolithography (STL) files or other computer-aided design files or geometric code"; Sections 3273.633 and 3273.634 set the September 1, 2028 algorithm performance-standard date, direct periodic review of emerging techniques "including, but not limited to ... volumetric search functionality," and address "acceptable false-positive and false-negative rates"; Section 3273.635 sets the March 1, 2029 manufacturer self-attestation deadline (a false attestation exposing the filer to perjury prosecution under Penal Code Section 118) and the June 1, 2029 published-list date, updated at least quarterly; Section 3273.636 requires manufacturers to equip printers with firearm blocking technology, lists the exemptions, makes the sale prohibition operative December 1, 2029, and authorizes a civil penalty not to exceed twenty-five thousand dollars ($25,000) per violation; Penal Code Section 29187 makes it a misdemeanor to knowingly disable or circumvent the blocking technology with intent to manufacture firearms, or to sell a modified listed printer with intent to facilitate the unlawful manufacture of firearms.</ref>
<ref name="bill-history">{{Cite web |url=https://leginfo.legislature.ca.gov/faces/billHistoryClient.xhtml?bill_id=202520260AB2047 |title=AB-2047 Bill History |publisher=California Legislative Information |access-date=2026-06-23}} Records the May 26, 2026 third-reading passage "(Ayes 58. Noes 19.)" and the May 27, 2026 entry "In Senate. Read first time. To Com. on RLS. for assignment."</ref>
<ref name="toms-ca">{{Cite web |url=https://www.tomshardware.com/3d-printing/california-assembly-passes-3d-printer-bill-that-would-criminalize-bypassing-mandated-gun-blocking-software |title=California Assembly passes 3D printer bill that would criminalize bypassing mandated gun-blocking software |author=Luke James |publisher=Tom's Hardware |date=2026-05-30 |access-date=2026-06-23}} Reports Assembly passage and transfer to the Senate, the DOJ performance-standard and list dates, the $25,000 penalty, the open-firmware (Marlin, Klipper) exposure, the exemptions that omit consumer machines in schools, libraries, and makerspaces, and that detection "may exceed the compute available on consumer printers, pushing checks to remote servers."</ref>
<ref name="ap-lieb">{{Cite web |url=https://www.pbs.org/newshour/nation/first-of-its-kind-law-in-new-york-could-block-3d-printers-from-making-guns |title=First-of-its-kind law in New York could block 3D printers from making guns |author=David A. Lieb |publisher=Associated Press, via PBS NewsHour |date=2026-06-13 |access-date=2026-06-23}} Quotes Julian Chultarsky, "a technical account manager at Physna, a Columbus, Ohio-based company," saying "Geometric search is mature, it's deployed, it is ready to be applied to this problem"; describes Solomon Diamond, an associate engineering professor at Dartmouth, at the online seminar; and quotes Bill Decker, executive chairman of the Association of 3D Printing, saying "it's not going to work" and "It's more of a political statement than anything else."</ref>
<ref name="sbir-geom">{{Cite web |url=https://www.sbir.gov/awards/205940 |title=Geometric Deep Learning Technology for Supply Chain (SBIR Phase II, FA8649-23-P-0450) |publisher=U.S. Small Business Administration, SBIR.gov |date=2023 |access-date=2026-06-23}} Air Force SBIR Phase II award to Physna Inc., $1,040,000; abstract states "Physna has developed the world's first true 3D geometric search engine and deep learning platform" enabling "true 3D search by geometry, automated labeling of digital assets based on shape, retrieval of 3D models with 2D images and photographs."</ref>
<ref name="usaf-geom">{{Cite web |url=https://www.usaspending.gov/award/CONT_AWD_FA864923P0450_9700_-NONE-_-NONE-/ |title=Award FA864923P0450 to PHYSNA INC, "Geometric Deep Learning Technology for Supply Chain" |publisher=USAspending.gov |access-date=2026-06-23}} Department of the Air Force contract to PHYSNA INC, $1,040,000, period of performance February 21, 2023 to February 22, 2024. The award page is a dynamic application; its structured record (recipient, amount, description "GEOMETRIC DEEP LEARNING TECHNOLOGY FOR SUPPLY CHAIN") is served at https://api.usaspending.gov/api/v2/awards/CONT_AWD_FA864923P0450_9700_-NONE-_-NONE-/</ref>
<ref name="usaf-sbir3">{{Cite web |url=https://www.usaspending.gov/award/CONT_AWD_FA701425C0023_9700_-NONE-_-NONE-/ |title=Award FA701425C0023 to PHYSNA INC, "Physna SBIR III Technology of Supply" |publisher=USAspending.gov |access-date=2026-06-23}} Department of the Air Force contract to PHYSNA INC, $1,500,000, beginning March 25, 2025. Structured record (description "PHYSNA SBIR III TECHNOLOGY OF SUPPLY") at https://api.usaspending.gov/api/v2/awards/CONT_AWD_FA701425C0023_9700_-NONE-_-NONE-/</ref>
<ref name="usaf-multimodal">{{Cite web |url=https://www.usaspending.gov/award/CONT_AWD_FA810024P0001_9700_-NONE-_-NONE-/ |title=Award FA810024P0001 to PHYSNA INC, multimodal image search in support of USAF Digital Material Management |publisher=USAspending.gov |access-date=2026-06-23}} Department of the Air Force contract to PHYSNA INC, $1,900,000; description "SMALL BUSINESS INNOVATION RESEARCH, TACTICAL FUNDING INCREASE (TACFI) SEQUENTIAL PHASE II +, TOPIC NO. AFX234-DCSO2, MULTIMODAL IMAGE SEARCH CAPABILITIES IN SUPPORT OF USAF DIGITAL MATERIAL MANAGEMENT (DMM)." Structured record at https://api.usaspending.gov/api/v2/awards/CONT_AWD_FA810024P0001_9700_-NONE-_-NONE-/</ref>
<ref name="usaf-altsource">{{Cite web |url=https://www.usaspending.gov/award/CONT_AWD_FA701424P0043_9700_-NONE-_-NONE-/ |title=Award FA701424P0043 to PHYSNA INC, alternate sources of supply through comparison of digital files |publisher=USAspending.gov |access-date=2026-06-23}} Department of the Air Force contract to PHYSNA INC, $530,000; description "ONE-TIME SERVICE TO IDENTIFY ALTERNATE SOURCES OF SUPPLY THROUGH COMPARISON OF DIGITAL FILES." Structured record at https://api.usaspending.gov/api/v2/awards/CONT_AWD_FA701424P0043_9700_-NONE-_-NONE-/</ref>
<ref name="gsa">{{Cite web |url=https://www.usaspending.gov/award/CONT_IDV_47QTCA24D00BN_4732/ |title=Federal Supply Schedule contract 47QTCA24D00BN to PHYSNA INC |publisher=USAspending.gov |access-date=2026-06-23}} General Services Administration, Federal Acquisition Service, Federal Supply Schedule contract to PHYSNA INC, beginning June 25, 2024. Structured record at https://api.usaspending.gov/api/v2/awards/CONT_IDV_47QTCA24D00BN_4732/</ref>
<ref name="pat531">{{Cite web |url=https://patents.google.com/patent/US11422531B2/en |title=US Patent 11,422,531 B2, System and methods for 3D model evaluation |publisher=United States Patent and Trademark Office, via Google Patents |date=2022-08-23 |access-date=2026-06-23}} Granted August 23, 2022; assignee Physna Inc.; inventors Glenn Warner and Paul Powers.</ref>
<ref name="pat736">{{Cite web |url=https://patents.google.com/patent/US12147736B2/en |title=US Patent 12,147,736 B2, System and methods for 3D model evaluation using triangle mesh hashing |publisher=United States Patent and Trademark Office, via Google Patents |date=2024-11-19 |access-date=2026-06-23}} Granted November 19, 2024; assignee Physna Inc.; inventors Glenn Warner and Paul Powers; abstract describes neighbor-facet edge hashing of a triangle mesh into an object hash value and comparison against reference models, including detection of added or removed sub-parts.</ref>
<ref name="powers-founder">{{Cite web |url=https://techcrunch.com/2021/01/29/three-dimensional-search-engine-physna-wants-to-be-the-google-of-the-physical-world/ |title=Three-dimensional search engine Physna wants to be the Google of the physical world |author=Jonathan Shieber |publisher=TechCrunch |date=2021-01-29 |access-date=2026-06-23}} Identifies "Physna founder Paul Powers."</ref>
<ref name="eff-ca">{{Cite web |url=https://www.eff.org/deeplinks/2026/04/dangers-californias-legislation-censor-3d-printing |title=The Dangers of California's Legislation to Censor 3D Printing |author=Cliff Braun and Rory Mir |publisher=Electronic Frontier Foundation |date=2026-04-13 |access-date=2026-06-23}} Calls A.B. 2047 a mandate for "censorware" that is "doomed to fail for its intended purpose," states it would "criminalize the use of open-source alternatives," and warns it would "render all 3D printer users vulnerable to surveillance" and enable DRM-style platform lock-in.</ref>
<ref name="techdirt-ca">{{Cite web |url=https://www.techdirt.com/2026/04/27/californias-3d-printer-law-would-criminalize-open-source-enshittify-the-3d-printing-space/ |title=California's 3D Printer Law Would Criminalize Open Source, Enshittify The 3D Printing Space |author=Karl Bode |publisher=Techdirt |date=2026-04-27 |access-date=2026-06-23}} Reproduces Adafruit's Phillip Torrone describing firearm-blueprint detection as "a classification problem with enormous false positive and false negative rates" that would have to identify every firearm component from raw STL/GCODE files "while not flagging pipes, tubes, blocks, brackets, gears."</ref>
</references>