Metraipolis › archive › dc_5ef8f0ded185

Agents That Edit Documents: Measuring Agentic PDF Forgery Against a Non-Agentic Control — arXiv:2609.23953

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.23953
License (as archived)
CC0-1.0
Archived
2026-10-09T07:01:18.635Z
SHA-256 of the archived text
0c2abcf80ac042de2a12230c9378583e051db034a5629c0326022c1d4993f7f5
Stable URL of this page
https://metraipolis.com/documento/dc_5ef8f0ded185

How to cite

Agents That Edit Documents: Measuring Agentic PDF Forgery Against a Non-Agentic Control — arXiv:2609.23953 — arXiv (en), https://arxiv.org/abs/2609.23953 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_5ef8f0ded185 (Metraipolis archive dc_5ef8f0ded185, sha256 0c2abcf80ac042de2a12230c9378583e051db034a5629c0326022c1d4993f7f5)
BibTeX
@misc{dc5ef8f0ded185,
  title = {Agents That Edit Documents: Measuring Agentic PDF Forgery Against a Non-Agentic Control — arXiv:2609.23953},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.23953},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_5ef8f0ded185; sha256 0c2abcf80ac042de2a12230c9378583e051db034a5629c0326022c1d4993f7f5},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.23953. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

AI agents that carry a multi-step computer task through on their own became ordinary tools in the past year, and the same autonomy is available to anyone whose task is harmful. We ask what that means for a relying party -- an insurer, a lender, an auditor -- whose evidence is a filed PDF. AgentForge-Bench measures how reliably an off-the-shelf coding agent, driving one of seven open-weight models with a shell and the stock Python PDF stack, alters one dollar amount, date or address in a real filed financial document from a single sentence of intent, graded by rules rather than by a model. Across 1,750 cells, 1,419 (81.1%) satisfy the verifier, and 808 (46.2%) also survive every stricter filter: visible, localized, typeface-matched, original value gone document-wide. A deterministic script with no model in it solves 98 of the 125 documents; the agents solve 124, and none the script solves alone. Agents misreport 41% of their wrong edits as done, no model refused, and the cheapest verified forgery costs 2.4 cents. The raw rate overstates the threat by about a factor of two; the strict rate is still large.

Tema degli esploratori: fiducia digitale e crittografia (retro-catalogo arXiv: annunci del 2026-09-22 in cs.CR).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Cryptography and Security (cs.CR).
Pagina del lavoro: https://arxiv.org/abs/2609.23953 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-09T06:56:01.031Z dalla pagina https://arxiv.org/catchup/cs.CR/2026-09-22?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).