AIToday
Large Language ModelsAI Regulation & PolicyAmazon AI BlogPublished: Jul 2, 2026, 04:00 JST1 min read

Inscribe detects document fraud in 90 seconds using AI on AWS

Inscribe detects document fraud in 90 seconds using AI on AWS

Key takeaway

  • Inscribe, a document fraud detection vendor serving banks and fintechs since 2017, deployed an agentic AI system on Amazon Bedrock that analyzes financial documents in under 90 seconds, 20 times faster than manual review.

  • The system coordinates multiple AI models to detect deepfakes, fabricated documents, and coordinated fraud while generating audit-ready reports—addressing a scale where fraud appears in 1 of every 16 documents and AI-generated forgeries grew 5x in late 2025.

3 Key Points

  1. What happened

    Inscribe built an AI system using Amazon Bedrock that detects tampered, fabricated, and AI-generated financial documents in under 90 seconds—a 20x improvement over traditional manual review, which takes 30 minutes per application.

  2. Why it matters

    Financial institutions processing thousands of loan and credit applications daily face fraud in 1 of every 16 documents, with AI-generated forgeries growing 5x from April to December 2025. Manual review cannot keep pace with volume or detect sophisticated deepfakes and coordinated fraud rings, leaving institutions exposed to millions in losses and regulatory risk.

  3. What to watch

    Inscribe uses multiple models from Amazon Bedrock matched to specific tasks—Claude Haiku 4.5 for routine document parsing (40% cost reduction vs. Claude Sonnet), Meta Llama models for transaction analysis, and Claude Sonnet 4 and 4.5 for cross-document fraud pattern detection and audit-ready reporting.

Ask the AI about this article →

FAQ

How much faster is Inscribe's system than manual review?
Inscribe detects fraud in under 90 seconds, compared with 30 minutes for traditional manual review—a 20x improvement.
Which AI models does Inscribe use and why?
Inscribe uses Claude Haiku 4.5 for routine document parsing and field extraction (achieving 40% cost reduction vs. Claude Sonnet), Meta Llama models for transaction analysis, and Claude Sonnet 4 and 4.5 for cross-document fraud analysis and report generation. The approach allows Inscribe to match each model to the task best suited to it.
How common is document fraud according to the data cited?
Fraud appears in 1 of every 16 documents, and AI-generated forgeries grew 5x from April to December 2025, according to Inscribe's 2026 State of Document Fraud Report.
Amazon AI BlogRead Original Article

Get the latest Large Language Models news every morning

For example, today's edition would include:

  • CBTS launches Forge Agents for custom AI agentsSiliconANGLE AI · 46m ago
  • Imec CEO: AI era widens chip-model-CSP collaborationDIGITIMES Asia · 46m ago
  • Alphabet's AI Overviews reach 2.5B monthly usersYahoo Finance AI · 46m ago

AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.

Free · takes 30 seconds · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. Q&As are published on this page for other readers too.

Related Articles

Next articleGoogle, NYC schools convene AI summit to bridge hiring and education gap