Prompt injection in immigration and asylum documentation AI

Four adversarial injection surfaces in immigration and asylum documentation AI

1. Asylum claim evidence image injection (INA §208, 8 CFR §208.13, UNHCR Credibility Guidelines)

Asylum claim evidence AI processes country condition evidence document scan images, asylum applicant identification document scan images, persecution evidence photograph images, country condition report page images from the State Department Country Reports on Human Rights Practices and the UNHCR country guidance documents, corroborating witness statement document images, documentary evidence of past harm or well-founded fear of future persecution images, and asylum applicant credibility-supporting biographical document images submitted through USCIS ELIS AI-assisted adjudication pipelines at 8 million or more immigration applications per year and the $4.4 billion annual USCIS adjudication workload; LexisNexis immigration AI at 400,000 or more legal professional users processing asylum research, case prediction, and country condition document analysis through AI-assisted legal research and credibility assessment tools; and Boundless AI at 200,000 or more immigration applications assisted through AI document review and form completion tools that process asylum supporting document scan images for USCIS credible fear determination and asylum petition credibility assessment input — extracting credible fear determination classifications and asylum petition credibility assessments from asylum claim country condition evidence and persecution evidence document image inputs in AI-assisted USCIS asylum adjudication pipelines at processing volumes that make individual human adjudicator examination of every AI-processed asylum evidence image impracticable across the USCIS asylum office and asylum interview preparation workloads.

The adversarial injection surface is the asylum claim evidence document scan or persecution evidence photograph image submission pathway: USCIS ELIS AI or LexisNexis immigration AI asylum evidence document image inputs submitted through AI-assisted credible fear determination and asylum petition credibility assessment tools for AI credibility classification record generation and USCIS asylum adjudication decision input. An adversarially crafted asylum evidence image — in which pixel perturbations applied to the document timestamp and date indicator display region of a persecution evidence photograph, the country condition document authenticity signal region of a supporting document image, or the identification document security feature and consistency indicator display in an asylum applicant identity document scan image cause the AI to classify fabricated persecution evidence as authentic corroborating evidence not meeting USCIS credibility challenge criteria when the actual document images evidence document fabrication indicators, timestamp inconsistencies, or security feature anomalies that would otherwise generate an asylum adjudicator credibility challenge referral — can suppress a document authenticity indicator that would otherwise generate an asylum claim denial recommendation, a credibility finding adverse to the applicant, and an asylum fraud referral record. In USCIS asylum office operations where ELIS AI processes asylum claim evidence document images to support asylum officer interview preparation and credibility assessment without individual human forensic examination of every AI-processed evidence image at the pixel level before the AI classification governs the adjudication recommendation, adversarial suppression of asylum evidence authenticity and fabrication indicators creates INA §208, 8 CFR §208.13, and 18 USC §1546 immigration document fraud dimensions of significant national security and refugee protection integrity severity.

The INA §208, 8 CFR §208.13, UNHCR credibility guideline, and 18 USC §1546 consequences of adversarially suppressed asylum evidence authenticity classification span INA §208 (8 USC §1158) asylum eligibility requirements establishing that an applicant who has been persecuted or has a well-founded fear of persecution on account of race, religion, nationality, membership in a particular social group, or political opinion may be granted asylum; 8 CFR §208.13 asylum approval criteria establishing that the burden of proof is on the applicant to establish eligibility for asylum by a preponderance of the evidence including corroborating evidence of past persecution or well-founded fear; UNHCR Handbook on Procedures and Criteria for Determining Refugee Status paragraphs 195–205 credibility assessment guidelines establishing that adjudicators should give the benefit of the doubt to applicants whose testimony is consistent and plausible even where individual documentary corroboration is incomplete; and 18 USC §1546 fraud and misuse of visas, permits, and other entry documents creating criminal penalties of imprisonment up to 10 years for any person who knowingly forges, counterfeits, alters, or falsely makes any document required under immigration laws or uses any such document knowing it to be forged, counterfeit, altered, or falsely made. Adversarial manipulation of asylum evidence AI that suppresses persecution photograph fabrication indicators or country condition document authenticity anomaly signals creates asylum adjudication integrity failures at both the national security dimension — enabling fraudulent asylum claims from applicants who are not genuine refugees — and the refugee protection failure dimension — potentially enabling adverse credibility findings against genuine refugees whose authentic evidence was corrupted by adversarial suppression of corroboration signals. Threshold: 55 for asylum claim evidence AI — a deliberately low threshold reflecting the dual consequences of false negatives in this context: national security and refugee protection failure.

2. Immigration petition document injection (INA §204, 18 USC §1001, 18 USC §1546)

Immigration petition document AI processes Form I-130 family petition supporting document scan images including birth certificate images, marriage certificate images, adoption decree images, and family relationship evidence photographs; Form I-140 employment-based immigrant petition supporting document scan images including foreign educational credential and transcript images, labor certification (PERM) application approval notice images, specialty occupation job offer letter images, and employer financial statement images; Form I-485 adjustment of status supporting evidence images including medical examination report images, civil document birth and marriage certificate images, prior immigration status documentation images, and financial sponsorship Form I-864 affidavit of support supporting document images — submitted through USCIS ELIS AI-assisted petition document review pipelines at 8 million or more annual application processing volumes and the $4.4 billion USCIS adjudication workload; Boundless AI at 200,000 or more immigration applications assisted through AI document completeness review and form preparation tools that process family petition and employment petition supporting document scan images for USCIS submission accuracy; and LexisNexis immigration AI at 400,000 or more legal professional users processing immigration petition document analysis and case management through AI-assisted legal research and document review tools — extracting petition document completeness, authenticity, and consistency assessments from Form I-130, I-140, and I-485 supporting document scan images in AI-assisted USCIS immigration petition adjudication pipelines.

The adversarial injection surface is the Form I-130 birth certificate or marriage certificate scan image submission pathway, or the Form I-140 foreign educational credential or PERM labor certification scan image submission pathway: USCIS ELIS AI or Boundless AI immigration petition supporting document scan images submitted through AI-assisted petition document authenticity and completeness review tools for AI petition document accuracy classification record generation and USCIS adjudication officer review input. An adversarially crafted Form I-130 birth certificate scan image or Form I-140 foreign educational credential image — in which pixel perturbations applied to the document stamp authenticity indicator display region of a birth certificate or educational credential image, the font consistency and date format compliance visual marker of a marriage certificate scan, or the labor certification approval notice security feature indicator in a PERM application document image cause the AI to classify a forged or altered supporting document as an authentic document passing USCIS petition document authenticity review criteria when the actual document evidences stamp inconsistencies, font irregularities, date format violations, or security feature anomalies that would otherwise generate a USCIS Request for Evidence (RFE) referral, a petition denial for lack of credible evidence, and a document fraud investigation referral — can suppress a petition document forgery indicator that would generate a material evidence integrity challenge. In USCIS petition adjudication operations where ELIS AI processes thousands of Form I-130, I-140, and I-485 supporting document scan images per day without individual human adjudicator pixel-level forensic examination of every AI-processed petition document before the AI classification governs the petition acceptance or RFE issuance decision, adversarial suppression of petition supporting document authenticity indicators creates INA §204, 18 USC §1001, and 18 USC §1546 immigration petition fraud dimensions.

The INA §204, 18 USC §1001, 18 USC §1546, and USCIS document fraud enforcement consequences of adversarially suppressed immigration petition document authenticity classification span INA §204 (8 USC §1154) petition requirements establishing the evidentiary standards for family-based and employment-based immigrant visa petitions; 18 USC §1001 false statements to government agencies creating criminal penalties of up to $500,000 fine and five years imprisonment (eight years if the offense involves terrorism) for knowingly and wilfully making any materially false, fictitious, or fraudulent statement or representation in any matter within the jurisdiction of the executive branch of the government of the United States including USCIS petition proceedings; 18 USC §1546(a) fraud and misuse of visas creating criminal penalties of imprisonment up to 10 years for any person who knowingly makes any false statement with respect to a material fact in any application, affidavit, or other document required by the immigration laws or regulations; and USCIS administrative document fraud enforcement authority under INA §274C (8 USC §1324c) creating civil penalties of $250 to $2,000 per fraudulent document for first violations and $2,000 to $5,000 per document for subsequent violations, with referral authority to the Department of Justice for criminal prosecution of egregious or systematic petition document fraud patterns. Threshold: 55 for immigration petition document AI — reflecting INA §204 petition integrity requirements, 18 USC §1001 false statements criminal penalties, 18 USC §1546 immigration document fraud imprisonment, and USCIS §274C civil penalty dimensions.

3. Travel document image injection (1951 Refugee Convention, INA §101(a)(42), 18 USC §1546)

Travel document AI processes passport biographic page scan images including MRZ (machine-readable zone) data strip images, biographical data page security feature images, and laser-engraved personalization layer images; visa stamp document scan images including consular visa sticker images and visa validity indicator images; I-94 arrival and departure record images processed at CBP ports of entry; refugee travel document scan images issued under 1951 Refugee Convention Article 27 obligations to recognized refugees; advance parole travel document images; and border crossing card biometric and document authentication images — processed by NEC Corporation and Unisys CBP document AI at 300 or more US ports of entry screening 1 million or more daily international travelers through AI-assisted travel document authenticity analysis and biometric identity verification tools that compare live biometric capture with stored biometric data in CBP IDENT and DHS HART biometric identity systems; Tyler Technologies immigration case AI at 20,000 or more government clients including county immigration courts and state justice system operations processing travel document scan images through AI-assisted immigration case management tools; and USCIS ELIS AI at admission status verification and adjustment of status proceedings processing lawful admission documentation images for Form I-485 immigration benefit eligibility assessment.

The adversarial injection surface is the passport biographic page scan image or visa stamp document scan image submission pathway at CBP automated passport control kiosks and officer workstations: NEC/Unisys CBP document AI passport biographic page scan images and visa stamp document images submitted through AI-assisted travel document authenticity and biometric identity matching tools for AI travel document acceptance or referral classification record generation and CBP officer secondary inspection decision input. An adversarially crafted passport biographic page scan image or visa stamp document image — in which pixel perturbations applied to the MRZ checksum indicator display region of a machine-readable passport page image, the holographic security overlay consistency visual marker of a visa sticker scan image, or the biometric chip data integrity indicator in a biometric passport document image cause the AI to classify a counterfeit or altered travel document as a genuine document not meeting CBP referral for secondary inspection criteria when the actual travel document scan evidences MRZ checksum anomaly, security feature absence, or biometric data inconsistency indicators that would otherwise generate a CBP secondary inspection referral, an inadmissibility determination, and a document fraud seizure and referral record — can suppress a travel document forgery indicator that would otherwise generate a border security response. In CBP automated passport control operations at major international airports and land border crossings where NEC/Unisys CBP document AI processes over 1 million daily traveler document images without individual human CBP officer pixel-level forensic examination of every AI-processed travel document before the AI classification governs the traveler processing recommendation, adversarial suppression of passport and visa document security feature and biometric consistency indicators creates 18 USC §1543, 18 USC §1546, and INA §212(a)(6)(C) inadmissibility ground dimensions of significant border security severity.

The 1951 Refugee Convention Article 27, INA §101(a)(42), 18 USC §1543, 18 USC §1546, and CBP enforcement consequences of adversarially suppressed travel document authenticity classification span 1951 Refugee Convention Article 27 travel document obligations requiring contracting states to issue travel documents to recognized refugees lawfully residing in their territory who do not possess a valid national passport; INA §101(a)(42) (8 USC §1101(a)(42)) refugee definition establishing persecution-based refugee eligibility applicable to refugee travel document holders processed at CBP ports of entry; 18 USC §1543 forgery of passports and alteration of passports creating criminal penalties of imprisonment up to 25 years for forgery, counterfeiting, alteration, or false making of any passport or passport card, with an enhanced 25-year maximum for offenses committed in furtherance of drug trafficking or terrorism under 18 USC §1543(b); 18 USC §1546 fraud and misuse of visas creating criminal penalties of imprisonment up to 10 years for any person who knowingly uses any visa, permit, border crossing card, alien registration receipt card, or other document required for entry into the United States knowing the document to be forged, counterfeit, altered, or falsely made, with an enhanced 25-year maximum for terrorism-related offenses; and CBP administrative enforcement authority under INA §212(a)(6)(C) (8 USC §1182(a)(6)(C)) creating inadmissibility grounds for any alien who procures or attempts to procure admission by fraud or willful misrepresentation. Threshold: 50 for travel document AI at CBP ports of entry — the most conservative threshold in the immigration context, reflecting 18 USC §1543 forgery 25-year imprisonment, 18 USC §1546 terrorism nexus 25-year imprisonment, CBP port-of-entry border security criticality, and the irreversibility of a failed detection at an international border crossing.

4. Immigration court hearing evidence injection (8 CFR §1240.9, EOIR evidence rules)

Immigration court hearing evidence AI processes hearing evidence document scan images submitted in EOIR removal proceedings under INA §240, country condition report page images from State Department Country Reports and UNHCR country guidance documents submitted as corroborating evidence, expert witness report document scan images submitted to support asylum, withholding of removal, and Convention Against Torture (CAT) claims, corroborating evidence document scan images including medical records, police reports, news articles, and affidavit statement images, and legal brief and motion document images submitted by immigration counsel — processed by EOIR immigration court AI across 3.5 million or more pending immigration court cases through AI-assisted case management and hearing evidence classification tools; Tyler Technologies immigration case AI at 20,000 or more government clients including county immigration courts processing hearing evidence images through AI-assisted case management tools; and LexisNexis immigration AI at 400,000 or more legal professionals globally processing immigration court hearing evidence analysis and case strategy through AI-assisted legal research and document review tools that inform hearing evidence submission strategy and evidence classification for immigration judge review. The volume of pending EOIR cases — exceeding 3.5 million as of 2026 — means that AI-assisted case management and evidence classification tools are processing hearing evidence document images at volumes that create AI pre-classification dependency across many immigration court hearing preparation workflows.

The adversarial injection surface is the hearing evidence document scan image or country condition report page image submission pathway: EOIR immigration court AI or Tyler Technologies case management AI hearing evidence document images submitted through AI-assisted evidence classification and case management tools for AI hearing evidence categorization record generation and immigration judge review preparation input. An adversarially crafted hearing evidence document scan image — in which pixel perturbations applied to the documentary corroboration consistency indicator display region of a country condition report page image, the expert witness report credential and authority visual marker of an expert declaration scan image, or the medical evidence and physical harm documentation indicator display in a medical record scan image submitted to support a CAT claim cause the AI to alter the evidence classification, suppress adverse evidence relevance signals, or cause AI case management tools to incorrectly categorize submitted hearing evidence documents — can cause the AI-assisted evidence summary presented to immigration court hearing officers to misrepresent the quality, consistency, or classification of submitted hearing evidence at a level of abstraction that influences immigration judge review and hearing outcome. In EOIR immigration court case management operations where AI tools process hearing evidence document images across the 3.5 million or more pending case docket without individual human immigration judge pixel-level examination of every AI-processed evidence document before the AI classification influences the hearing preparation and evidence summary presented for judicial review, adversarial manipulation of hearing evidence AI classification creates 8 CFR §1240.9 immigration court hearing procedure integrity and INA §240(b)(4)(B) due process dimensions.

The 8 CFR §1240.9, EOIR Practice Manual, INA §240(b)(4)(B), and due process consequences of adversarially manipulated immigration court hearing evidence classification span 8 CFR §1240.9 immigration judge hearing procedures establishing the procedural framework for evidence submission, record of proceeding management, and evidence authentication in EOIR removal proceedings; EOIR Practice Manual §3.1 evidence submission requirements establishing the standards for documentary evidence authentication, translation, and submission timing in immigration court proceedings; INA §240(b)(4)(B) (8 USC §1229a(b)(4)(B)) statutory right to present evidence in removal proceedings establishing that the alien shall have a reasonable opportunity to examine the evidence against the alien, to present evidence on the alien’s own behalf, and to cross-examine witnesses presented by the government; and the Fifth Amendment due process requirements applicable to EOIR removal proceedings established by the Supreme Court in Mathews v. Eldridge and its immigration-specific applications requiring that removal proceedings satisfy procedural due process standards. Adversarial manipulation of hearing evidence AI that alters the classification or suppresses the relevance signals of asylum corroborating evidence, CAT medical evidence, or withholding of removal country condition evidence in EOIR case management AI creates the conditions for due process violations in removal proceedings where AI-generated evidence summaries influence the completeness and accuracy of the hearing record presented for immigration judge review. Threshold: 60 for immigration court hearing evidence AI — reflecting 8 CFR §1240.9 hearing procedure integrity, EOIR Practice Manual evidence submission requirements, INA §240(b)(4)(B) due process rights in removal proceedings, and the irreversibility of wrongful removal orders as a consequence of adversarially corrupted hearing evidence classification.

Integration: immigration and asylum documentation AI image ingestion with Glyphward pre-scan

Immigration and asylum documentation AI image ingestion flows from USCIS ELIS AI asylum claim evidence document scan and persecution photograph image channels, ELIS AI and Boundless AI immigration petition Form I-130/I-140/I-485 supporting document scan ingestion interfaces, NEC/Unisys CBP document AI passport biographic page and visa stamp scan image processing systems at US ports of entry, and EOIR AI and Tyler Technologies case management AI immigration court hearing evidence document scan processing platforms into asylum claim credibility assessment AI, immigration petition document authenticity verification AI, travel document security feature and biometric consistency verification AI, and immigration court hearing evidence classification AI pipelines. Insert Glyphward’s pre-scan at the ingestion boundary before AI-generated output is committed to USCIS asylum adjudication records, immigration petition acceptance or RFE determination records, CBP port-of-entry processing determinations, or EOIR hearing evidence classification records:

import asyncio
import base64
import hashlib
import os
import uuid
from enum import Enum
from pathlib import Path

import httpx

GLYPHWARD_API_KEY = os.environ["GLYPHWARD_API_KEY"]
GLYPHWARD_SCAN_URL = "https://glyphward.com/v1/scan"

# Immigration & asylum documentation AI — INA §208, 8 CFR §208.13, UNHCR credibility guidelines;
# INA §204, 18 USC §1001 ($500k/5yr), 18 USC §1546 document fraud (10yr/25yr terrorism nexus);
# 18 USC §1543 passport forgery (25yr), CBP port-of-entry biometric identity;
# 8 CFR §1240.9 EOIR hearing procedures, INA §240(b)(4)(B) due process removal proceedings.
THRESHOLD_ASYLUM_CLAIM_EVIDENCE_AI        = 55  # ELIS/LexisNexis; INA §208; §208.13; §1546 10yr
THRESHOLD_IMMIGRATION_PETITION_DOCUMENT_AI = 55  # ELIS/Boundless; INA §204; §1001 $500k/5yr; §1546
THRESHOLD_TRAVEL_DOCUMENT_AI              = 50  # NEC/Unisys CBP; §1543 25yr; §1546 terrorism 25yr
THRESHOLD_IMMIGRATION_COURT_EVIDENCE_AI   = 60  # EOIR/Tyler Tech; 8 CFR §1240.9; §240(b)(4)(B)


class ImmigrationAsylumDocumentationAIContext(str, Enum):
    ASYLUM_CLAIM_EVIDENCE_AI         = "asylum_claim_evidence_ai"          # ELIS, LexisNexis, Boundless
    IMMIGRATION_PETITION_DOCUMENT_AI = "immigration_petition_document_ai"  # ELIS, Boundless, LexisNexis
    TRAVEL_DOCUMENT_AI               = "travel_document_ai"                # NEC/Unisys CBP, ELIS APC
    IMMIGRATION_COURT_EVIDENCE_AI    = "immigration_court_evidence_ai"     # EOIR, Tyler Technologies


def threshold_for(context: ImmigrationAsylumDocumentationAIContext) -> int:
    mapping = {
        ImmigrationAsylumDocumentationAIContext.ASYLUM_CLAIM_EVIDENCE_AI:         THRESHOLD_ASYLUM_CLAIM_EVIDENCE_AI,
        ImmigrationAsylumDocumentationAIContext.IMMIGRATION_PETITION_DOCUMENT_AI: THRESHOLD_IMMIGRATION_PETITION_DOCUMENT_AI,
        ImmigrationAsylumDocumentationAIContext.TRAVEL_DOCUMENT_AI:               THRESHOLD_TRAVEL_DOCUMENT_AI,
        ImmigrationAsylumDocumentationAIContext.IMMIGRATION_COURT_EVIDENCE_AI:    THRESHOLD_IMMIGRATION_COURT_EVIDENCE_AI,
    }
    return mapping[context]


async def scan_immigration_asylum_documentation_ai_image(
    image_path: str | Path,
    context: ImmigrationAsylumDocumentationAIContext,
    case_ref_hash: str,       # SHA-256 of A-number, receipt number, case ID, or hearing docket number
    document_ref: str,        # e.g. "ELIS-I130-2026-44821", "CBP-POE-JFK-2026-0041", "EOIR-2026-8819"
    adjudication_session_id: str,  # USCIS interview session, CBP processing session, or EOIR docket
    client: httpx.AsyncClient,
) -> dict:
    """
    Scan an immigration or asylum documentation AI image for adversarial injection payloads
    before forwarding to asylum claim credibility assessment, immigration petition document
    authenticity verification, travel document security feature and biometric identity
    verification, or immigration court hearing evidence classification AI.

    Raises AdversarialImmigrationAsylumDocumentationAIImageError if score meets threshold:
      - ASYLUM_CLAIM_EVIDENCE_AI:         threshold 55; INA §208; 8 CFR §208.13; 18 USC §1546 10yr
      - IMMIGRATION_PETITION_DOCUMENT_AI: threshold 55; INA §204; 18 USC §1001 $500k/5yr; §1546
      - TRAVEL_DOCUMENT_AI:               threshold 50; 18 USC §1543 forgery 25yr; §1546 terrorism 25yr
      - IMMIGRATION_COURT_EVIDENCE_AI:    threshold 60; 8 CFR §1240.9; INA §240(b)(4)(B) due process
    """
    image_bytes          = Path(image_path).read_bytes()
    image_b64            = base64.b64encode(image_bytes).decode()
    image_sha256         = hashlib.sha256(image_bytes).hexdigest()
    client_scan_id       = str(uuid.uuid4())
    threshold            = threshold_for(context)

    resp = await client.post(
        GLYPHWARD_SCAN_URL,
        headers={"Authorization": f"Bearer {GLYPHWARD_API_KEY}"},
        json={
            "image": image_b64,
            "source": context.value,
            "metadata": {
                "immigration_asylum_context": context.value,
                "case_ref_hash":             case_ref_hash,
                "document_ref":              document_ref,
                "adjudication_session_id":   adjudication_session_id,
                "client_scan_id":            client_scan_id,
                "image_sha256":              image_sha256,
            },
        },
        timeout=8.0,
    )
    resp.raise_for_status()
    result = resp.json()

    audit_record = {
        "case_ref_hash":             case_ref_hash,
        "document_ref":              document_ref,
        "adjudication_session_id":   adjudication_session_id,
        "immigration_asylum_context": context.value,
        "scan_id":                   result["scan_id"],
        "client_scan_id":            client_scan_id,
        "image_sha256":              image_sha256,
        "score":                     result["score"],
        "flagged_region":            result.get("flagged_region"),
        "threshold":                 threshold,
        "action":                    "blocked" if result["score"] >= threshold else "allowed",
    }
    await write_immigration_asylum_audit_record(audit_record)

    if result["score"] >= threshold:
        raise AdversarialImmigrationAsylumDocumentationAIImageError(
            f"Immigration asylum documentation AI image blocked [{context.value}]: "
            f"scan_id={result['scan_id']} score={result['score']} "
            f"case={case_ref_hash} doc={document_ref}"
        )
    return result


async def write_immigration_asylum_audit_record(record: dict) -> None:
    """Persist audit record to immigration adjudication regulatory documentation store (stub)."""
    import json, sys
    print(json.dumps(record), file=sys.stderr)


class AdversarialImmigrationAsylumDocumentationAIImageError(Exception):
    """Raised when an immigration or asylum documentation AI image exceeds the adversarial injection threshold."""
    pass

Call scan_immigration_asylum_documentation_ai_image() with ImmigrationAsylumDocumentationAIContext.ASYLUM_CLAIM_EVIDENCE_AI before forwarding USCIS ELIS AI or LexisNexis immigration AI asylum claim country condition evidence document images, persecution evidence photographs, and asylum applicant identification document scans to credible fear determination and asylum petition credibility assessment AI — with case_ref_hash as the SHA-256 of the USCIS A-number or receipt number for INA §208 asylum adjudication integrity, 8 CFR §208.13 credibility assessment accuracy, UNHCR credibility guideline compliance, and 18 USC §1546 immigration document fraud prevention audit trail documentation. Call with ImmigrationAsylumDocumentationAIContext.IMMIGRATION_PETITION_DOCUMENT_AI for USCIS ELIS AI or Boundless AI Form I-130 birth certificate and marriage certificate scan images, Form I-140 educational credential and PERM labor certification scan images, and Form I-485 supporting evidence scan images before petition document authenticity verification AI, with document_ref as the USCIS receipt number for INA §204 petition integrity, 18 USC §1001 false statement criminal penalty compliance, 18 USC §1546 immigration document fraud prevention, and USCIS §274C civil penalty audit trail. Call with ImmigrationAsylumDocumentationAIContext.TRAVEL_DOCUMENT_AI for NEC/Unisys CBP document AI passport biographic page scan images, visa stamp document images, and refugee travel document images at US ports of entry before travel document authenticity and biometric identity verification AI — with threshold 50, the lowest threshold in this product reflecting 18 USC §1543 passport forgery 25-year imprisonment and 18 USC §1546 terrorism nexus 25-year imprisonment at 300+ US ports of entry screening 1 million or more daily travelers. Call with ImmigrationAsylumDocumentationAIContext.IMMIGRATION_COURT_EVIDENCE_AI for EOIR AI or Tyler Technologies case management AI hearing evidence document scan images, country condition report page images, and expert witness report images before immigration court evidence classification AI, with adjudication_session_id for 8 CFR §1240.9 hearing procedure integrity, EOIR Practice Manual §3.1 evidence submission compliance, and INA §240(b)(4)(B) due process rights in removal proceedings audit trail documentation. Get early access

Coverage matrix

Related questions

Further reading

• FigStep adversarial image injection detection — technical documentation of FigStep pixel-level image injection attacks and Glyphward’s detection approach, directly applicable to asylum claim evidence photograph adversarial pixel perturbation and travel document biographic page adversarial manipulation in immigration document AI contexts.
• PDF prompt injection detection — related adversarial attack surface covering AI pipelines that process PDF document images, applicable to USCIS ELIS Form I-130 birth certificate PDF scans, Form I-140 educational credential PDF documents, and EOIR hearing evidence PDF document submissions that generate document page images for AI analysis.
• Vision-language model security — technical framework for adversarial injection attacks targeting vision-language model (VLM) image analysis pipelines, applicable to USCIS ELIS AI, LexisNexis immigration AI, and NEC/Unisys CBP document AI that use VLM architectures to analyze travel document images and asylum evidence photographs.
• Free tier — 10 scans/day, no card required — start scanning immigration and asylum documentation AI inputs at development volumes before committing to a production plan; threshold 50 for travel document AI at CBP ports of entry means every scan matters.
• Prompt injection scanner for document AI — integration guide for document AI platforms that process multi-page document scan images, applicable to USCIS ELIS AI petition document review, Boundless AI form completion document processing, and Tyler Technologies immigration case management AI that analyze Form I-130, I-140, and I-485 supporting document images.
• OWASP LLM01 prompt injection — multimodal — OWASP Top 10 for LLM Applications LLM01 prompt injection threat documentation with multimodal AI attack surface analysis applicable to immigration document AI vision model ingestion pipelines and the OWASP-defined injection threat categories relevant to USCIS, EOIR, and CBP document AI compliance frameworks.