Conference 14–15 May 2024 📍 Berlin, Germany

PDF Days Europe 2024 — PDF/UA-2 Arrives and AI Enters the Room

PDF Days Europe 2024 took place in Berlin on 14–15 May 2024, with PDF/UA-2 (ISO 14289-2:2024) finally published and AI-assisted document processing emerging as a major discussion topic alongside the perennial themes of accessibility, archiving, and digital signatures.

Official event page →

Key takeaways

PDF/UA-2 (ISO 14289-2:2024) was published one month before the conference — the first edition to address a finalized standard. Validator implementations were near-complete; authoring tool support was partial.
Post-quantum cryptography was addressed as an immediate planning concern: NIST's PQC standards were imminent, and long-term signed PDFs required a transition roadmap to PQC-based signature algorithms.
LLM RAG deployment at scale demonstrated that PDF tagging quality (especially table structure) directly improved AI extraction accuracy — connecting accessibility standards to AI performance outcomes.
The conference confirmed the hybrid format as a permanent fixture with growing remote participation alongside stable in-person Berlin attendance.

PDF Days Europe 2024 was held in Berlin in May 2024, organised by the PDF Association. The 2024 conference had a specific historical context: PDF/UA-2 (ISO 14289-2:2024) had been published in April 2024 — approximately one month before the event. The conference was therefore able to address the newly published standard rather than a near-final draft, and sessions focused on what PDF/UA-2 publication actually meant for practitioners who had been following its development. Separately, in August 2024, NIST finalised its post-quantum cryptography standards (FIPS 203, 204, 205). While those standards were published after PDF Days Europe 2024, the conference addressed the post-quantum transition roadmap that would incorporate them — and several sessions reviewed the preparatory work that PDF tool vendors and PKI infrastructure providers needed to begin immediately.

PDF/UA-2 publication: what it means in practice

The publication of PDF/UA-2 (ISO 14289-2:2024) one month before the conference made the 2024 event a milestone. For the first time, sessions could address a finalized standard rather than a specification in development. What changed from PDF/UA-1 — Sessions provided practitioner-focused summaries of the changes most relevant for document producers and tool developers: The /Aside element for tangential content, /FENote for footnotes/endnotes, and /DocumentFragment for partial documents were now officially normative. The removal of the /B conformance level was confirmed — there is one PDF/UA-2 standard and no partial compliance option. MathML for mathematical content is normative rather than advisory. Validator status at launch — Both PAC and veraPDF had PDF/UA-2 support in beta or release-candidate status at publication time. Sessions compared the two validators on a set of test documents, identifying cases where their results diverged. The divergences were primarily in new requirements where the specification text allowed multiple interpretations — a normal first-year validator situation. The PDF Association maintained an issues tracker for validator consensus work. Authoring tool readiness — Adobe InDesign had partial PDF/UA-2 support in current versions (the new structure elements were not yet produced automatically). Microsoft Word had no PDF/UA-2-specific changes in its export. The practical implication: organisations targeting PDF/UA-2 compliance for new documents would need post-export remediation using Acrobat Pro or specialist tagging tools for any content using the new structure elements. Migration from PDF/UA-1 — For organisations with existing PDF/UA-1 compliant document libraries, there was no mandatory migration requirement. PDF/UA-1 documents remain valid. The question was whether new documents should target PDF/UA-1 (maximum toolchain support) or PDF/UA-2 (current standard). The conference consensus: new accessibility programme work should target PDF/UA-2 for new document creation, while legacy remediation programmes could continue against PDF/UA-1 in the near term.

Post-quantum cryptography and PDF signatures

The post-quantum transition was a sustained topic at PDF Days Europe 2024. NIST's post-quantum standardisation process had been ongoing since 2016; by 2024 the finalists were known and the standards were imminent (published August 2024). The threat — A cryptographically relevant quantum computer running Shor's algorithm could break RSA and ECDSA signatures in polynomial time — the same algorithms that underpin all current PDF digital signatures. A long-term signed PDF — a digitally signed contract intended to remain legally valid for 30 years — may need to be verifiable after quantum computers become practical. Harvest now, decrypt later — For encryption (PDF password protection), the threat is immediate: adversaries can collect encrypted PDFs now and decrypt them when quantum computers are available. For signatures, the threat is different: a signature created today could be forged retrospectively once the signing algorithm is broken. Documents relying on signature validity for legal enforceability in the 2030s–2040s are potentially at risk. NIST PQC algorithms — The three standardised algorithms (CRYSTALS-Kyber for key encapsulation, CRYSTALS-Dilithium and FALCON for signatures, SPHINCS+ as a backup) were reviewed in sessions. For PDF signatures, the relevant algorithms are Dilithium and FALCON (digital signatures). The ETSI PAdES standards would need to be updated to include PQC algorithm suites. PDF Association roadmap — The PDF Association presented its planned roadmap for incorporating PQC into ISO 32000 (PDF 2.0 errata/revision) and into PAdES. The timeline was several years — standards processes move slowly — but the groundwork was being laid. In the near term, organisations were advised to implement Long-Term Validation (LTV) signatures with embedded OCSP/CRL data, which at minimum ensures that current signature validity can be confirmed without relying on CA infrastructure that might be compromised.

LLM integration and PDF quality standards

By 2024, large language model integration with PDF documents had moved from evaluation to production deployment at scale. PDF Days Europe 2024 addressed the quality standards that PDF documents needed to meet for effective LLM processing: RAG (Retrieval-Augmented Generation) and PDF — The dominant enterprise LLM deployment pattern in 2024 was RAG: a language model retrieves relevant passages from a document corpus and uses them as context for answering queries. PDF extraction quality determines what the model actually retrieves. Sessions reviewed common extraction failures that degraded RAG performance: ligature handling (the 'fi' ligature not extracted as separate 'f' and 'i'), hyphenation artifacts, tables extracted as prose, and mathematical notation producing garbled text. Structured data extraction — Beyond passage retrieval, organisations were using LLMs to extract structured data from PDFs (table contents, key-value pairs, dates, amounts). Properly tagged PDFs with accurate table structure (/Table, /TR, /TH, /TD elements) produced dramatically better extraction accuracy than untagged PDFs where the table had to be inferred from position. PDF as LLM training data — A side discussion addressed the use of PDFs as LLM training data — a significant portion of the Common Crawl dataset that trains many foundation models is PDF text extracted from the web. Low-quality PDF text (garbled by poor font encoding, reverse-engineered from image PDFs, or misextracting columns) contaminates training data. The PDF quality standards community was increasingly aware that good PDF authoring practices had downstream effects on AI model quality.

Frequently asked questions

What is post-quantum cryptography and why does it affect PDF signatures?

Post-quantum cryptography refers to cryptographic algorithms that are believed to be secure against attacks by quantum computers, unlike current algorithms (RSA, ECDSA, ECDH) which can be broken by a sufficiently powerful quantum computer running Shor's algorithm. PDF digital signatures use RSA or ECDSA — the same algorithms at risk. NIST's post-quantum standards (CRYSTALS-Dilithium, FALCON, SPHINCS+, CRYSTALS-Kyber) provide quantum-resistant alternatives. For PDF signatures, the transition matters most for: long-term signed documents where signature validity must be maintained for decades; high-value legal and financial documents; and government documents with long retention requirements. The practical transition requires: (1) updated PKI infrastructure at CAs to issue PQC certificates; (2) updates to PAdES and PDF signing standards; (3) updates to PDF signing tool software. This is a multi-year process expected to complete in the late 2020s.

What is Long-Term Validation (LTV) in PDF signatures?

Long-Term Validation (LTV) is a PDF signature enhancement that embeds the verification data needed to validate a signature long after it was created — even after the signing certificate, CA certificates, and OCSP/CRL services have expired or been decommissioned. Without LTV, a PDF signature that was valid at signing time may be unverifiable years later because: the certificate chain is no longer available online; OCSP responders are offline; CRL distribution points are unreachable. With LTV, the PDF contains embedded: the complete certificate chain from the end-entity certificate to the trust anchor; OCSP response or CRL proving the certificate was valid at signing time; a trusted timestamp from a TSA proving the signature existed at a specific time. PAdES-LT and PAdES-LTA formats (defined in ETSI EN 319 132) specify LTV signature structures for PDF. Acrobat Pro creates LTV-enabled signatures when signing with a qualified certificate connected to a functioning OCSP service.

How does PDF/UA-2 affect documents created before 2024?

PDF/UA-2 does not retroactively invalidate documents created to PDF/UA-1. A document that was fully conformant with PDF/UA-1 (ISO 14289-1:2012) remains conformant with that standard — standards do not have expiry dates. In practice: (1) Legacy PDF/UA-1 documents do not need to be migrated. (2) New document creation for accessibility programmes should target PDF/UA-2. (3) Procurement requirements in government and regulated sectors will transition from requiring PDF/UA-1 to requiring PDF/UA-2 over a 2–5 year period. (4) PDF validators (PAC, veraPDF) continue to validate both PDF/UA-1 and PDF/UA-2, so organisations can continue using existing validation workflows while transitioning to PDF/UA-2 targets for new production.

Try it in your browser

The PDF tools and standards discussed at this event are available free — no account needed.

Sign PDF Edit PDF Annotate PDF Open editor