Publishing an invention to stop anyone else patenting it works on one condition: the disclosure has to have been available to the public before their filing date, and that has to be provable years later. Everything else about a defensive publication can be right and the strategy still fails on that point.
This page sets out what makes a defensive publication effective, why the date is the part that gets attacked, and what a dated record does and does not fix.
| Requirement | What it means | If it fails |
|---|---|---|
| Enabling disclosure | Detailed enough that a skilled person could work the invention | Not novelty-destroying, so the competitor patent stands |
| Publicly available | Made available to the public, which is the legal test | It is not prior art at all |
| Discoverable | Indexed where examiners actually search | The patent is granted anyway, and you invalidate it later at cost |
| Provable date | A date the examining division can be convinced is correct | The examiner stops relying on the disclosure |
The first three are decisions you make when you publish. The fourth is decided years later by somebody else, using whatever evidence still exists.
A journal article has a date because a publisher put one on it. A web page usually does not, in any sense that survives scrutiny. The EPO Guidelines for Examination put the difficulty plainly in the section on internet disclosures at G-IV, 7.5: not all pages say when they were published, sites are easily updated, and most keep no archive that would let anyone establish what was published and when.
The standard applied is the balance of probabilities, but the Guidelines are careful about what that means here. At G-IV, 7.5.2 they state that it is not sufficient for the publication date to be merely probable: the examining division must be convinced that it is correct. A date that looks plausible is not a date you can rely on.
What follows from that is the part worth understanding before you publish anywhere. Under G-IV, 7.5.3, if an applicant gives reasons to question the date of a disclosure cited against them, the examiner has to take those reasons into account, and if no longer convinced the disclosure forms part of the state of the art, must either produce further evidence or stop using it. Your defensive publication can leave the picture without anyone deciding it was false. Generic complaints about internet reliability get nowhere. Specific ones do the damage.
When a disclosure carries no date, or a date has been shown to be unreliable, G-IV, 7.5.4 sets out where examiners look next. Internet archiving services, principally the Wayback Machine, which the Guidelines treat as credible for what it does capture even though it is incomplete. Search engine indexing dates, which the Guidelines note will be later than the actual publication date because indexing takes time. Date information embedded in the page and recoverable with forensic tools.
Every one of those is an inference drawn after the fact by someone reconstructing what happened. They work often enough. They are not the same thing as a record made at the moment of publication by a party with no interest in the outcome, and they are conspicuously worse when what you need to prove is not just roughly when, but exactly what.
The office has drawn the same conclusion for its own purposes. G-IV, 7.5.6 now requires the EPO to capture and store internet disclosures in a form suitable to prove their availability to the public, so the evidence survives if the page later disappears, with B-X, 11.6 providing for that material to be released to parties on request. Anyone relying on an online disclosure eventually reaches the same position: the page will not necessarily be there when it is needed, so the proof has to be made at the time.
Deposit the file you are about to publish, on the day you publish it. It is hashed with SHA-256, the hash is timestamped under RFC 3161 by an accredited trust service provider, and the timestamped hash is anchored to the Bitcoin blockchain. If an applicant later argues your disclosure did not contain the feature in issue, or was quietly revised afterwards, the fingerprint settles it, and the timestamp came from a party with no stake in the argument.
With eIDAS qualification the timestamp carries the presumption under Article 41(2) that the date and time are accurate and the data intact, which shifts the burden to whoever disputes it. That is a different position from asking an examiner to infer a date from an archive snapshot. The distinction between a qualified timestamp and a blockchain anchor alone is set out in blockchain timestamps compared with qualified timestamps.
Two limits, both worth stating before anyone infers more than is there. Provlyn does not publish anything and does not create prior art. You have to publish the substance yourself, somewhere it will be found, and this page is not a substitute for doing that.
And a timestamp proves that a file existed in a given form at a given moment. It does not prove the file was available to the public at that moment, which is the actual legal test. Public availability is established by evidence about the publication itself. What a deposit removes is any argument about the contents and the moment, which is one half of the question rather than the whole of it. If you are filing rather than publishing, the evidence that matters is different, and is set out in the dates a patent filing does not record.
The record itself can be checked by anyone, which matters when the point of it is to survive a challenge from somebody who does not trust you. Run SHA-256 over the published file and compare, verify the timestamp with OpenSSL, and check the issuing authority against the European Commission's trusted lists. Our timestamp validator is open source and checks any provider's timestamp, not only ours, and independent verification sets out each check in full.
Record your first deposit →Publishing an enabling description of an invention so that it enters the public domain and becomes prior art, which prevents anyone else from patenting it and preserves your own freedom to operate. It is sometimes called a defensive disclosure. There is no official procedure and no register: it is a practice rather than a legal instrument. The historic mechanisms that did carry the name in US law are gone, since the USPTO Defensive Publication Program ended in the 1980s and Statutory Invention Registration was repealed by the America Invents Act with effect from 16 March 2013.
Three things, and only the first is about the document. It has to be enabling, meaning detailed enough that a skilled person could work the invention from it, which usually means the same content you would put in an application. It has to be somewhere an examiner will find it, which is why practitioners use established prior art databases and technical journals rather than a company blog. And its publication date has to survive being questioned, because a disclosure whose date cannot be established is a disclosure the examiner will stop relying on.
Somewhere indexed and searched by the people who need to find it. Established routes include commercial prior art databases such as IP.com and Research Disclosure, technical journals, and filing an application you allow to publish. In principle any publicly accessible medium can qualify, since the legal test is whether the matter was made available to the public rather than where it appeared. In practice the point of the exercise is deterrence, and a disclosure nobody finds deters nobody, so the venue matters more than the law strictly requires.
Because an internet disclosure has no inherent date. The EPO Guidelines for Examination address this directly at G-IV, 7.5, noting that not all web pages say when they were published, that sites are easily updated, and that most keep no archive letting anyone establish what was published and when. Under the standard of proof at G-IV, 7.5.2 it is not enough that the publication date is merely probable: the examining division has to be convinced it is correct. A date that only looks right is not sufficient.
The burden shifts. Under G-IV, 7.5.3 the examiner raising an objection must substantiate it, but if an applicant gives reasons to question the publication date those reasons must be taken into account, and if the examiner is no longer convinced the disclosure forms part of the state of the art, they must either produce further evidence or stop using it as prior art. Generic complaints about the reliability of the internet carry little weight. Specific reasons carry a great deal. Your defensive publication can drop out of the picture without anyone deciding it was untrue.
G-IV, 7.5.4 sets out the fallbacks. Records from an internet archiving service, principally the Wayback Machine, whose incompleteness the Guidelines say does not detract from the credibility of what it does capture. Indexing dates from search engines, which the Guidelines note will be later than the actual publication date because indexing takes time. Date information embedded in the page and retrievable with forensic tools. Each of these is an inference about when something appeared. None is a record made at the moment of publication by a party with no stake in the outcome.
Yes, and the practice was tightened recently. G-IV, 7.5.6 now requires the EPO to capture and store internet disclosures in a form suitable to prove availability to the public, so that the evidence survives if the disclosure later disappears from the internet, and B-X, 11.6 provides for that evidence to be made available to parties on request. The office reached the same conclusion anyone relying on an online disclosure eventually reaches: the page will not necessarily be there when it is needed, so the proof has to be made at the time.
It records what you published and when, so the two things most likely to be attacked are fixed. The published file is hashed with SHA-256, the hash is timestamped by an accredited trust service provider, and the result is anchored to the Bitcoin blockchain. If someone later argues your disclosure did not contain the feature at issue, or was edited after the fact, the fingerprint answers it. Provlyn does not publish anything and does not create prior art. You still have to publish, and where you publish still matters.
No, and the distinction matters because public availability is the legal test. A timestamp establishes that a file existed in a specific form at a specific time and has not changed since. It does not establish that the file was accessible to the public at that moment, which is a separate question answered by evidence about the publication itself, such as the venue, its access arrangements and its records. A timestamp makes one half of the question unarguable. It does not answer the other half.
That is a strategy question for a patent attorney, and it turns on whether you want to exclude others or merely stop being excluded yourself. Publishing gives no positive right: it prevents others patenting the disclosed matter but grants you nothing to enforce. It is also one-way, since publication destroys your own novelty in Europe, where there is no general grace period. Companies commonly patent selectively and publish the incremental improvements around it, but which improvements fall on which side is a judgement no page can make for you.
This page gives general information about defensive publication and evidence of publication dates. It is not legal advice, and nothing here is a recommendation about whether to publish or to file. References to the EPO Guidelines for Examination are to the sections named; the Guidelines are revised annually and section numbering can change. For a specific matter, consult a qualified patent attorney.