Museums, Libraries, and Cultural Heritage IP Toolkit: Collections, Digitisation, and Access

By Casey Scott McKay · 2024-06-20

Cultural institutions hold objects they own and rights they mostly do not, and the gap between those two positions is where every difficult decision in the sector sits. This toolkit collects the framework. It works the public domain determination that clears more material than institutions expect, the deed of gift audit that decides what an institution may actually do, and the diligent search that confers no permission and is still worth doing. It then covers section 108's on-premises limitation, the fair use position the library cases established, the accessibility route with two independent legal bases, the tiered access model, rights metadata, the takedown operation, community consultation, and the institution's own rights.

IP and Technology > Copyright | Toolkit | Published 20 June 2024 - Updated 13 April 2025 | Casey Scott McKay - marksy.us

Summary. Cultural institutions hold objects they own and rights they mostly do not, and the gap between those two positions is where every difficult decision in the sector sits. This toolkit collects the framework. It works the public domain determination that clears more material than institutions expect, the deed of gift audit that decides what an institution may actually do, and the diligent search that confers no permission and is still worth doing. It then covers section 108's on-premises limitation, the fair use position the library cases established, the accessibility route with two independent legal bases, the tiered access model, rights metadata, the takedown operation, community consultation, and the institution's own rights.

Keywords: cultural heritage IP toolkit · collections rights · deeds of gift · orphan works · diligent search · section 108 preservation · fair use for libraries · HathiTrust · public domain determination · renewal research · restored foreign works · Bridgeman · rights metadata · standardised rights statements · access tiers · takedown operations · community consultation · NAGPRA · image licensing · exhibition rights

Start Here

A curator asks whether the institution can put a collection online, and the answer requires separating two things that feel like one.

Physical custody is not copyright. The institution owns the object. Copyright in the work embodied in it belongs to its author, or to their heirs, or to a publisher, or to nobody who can be found.

Which means a collection of forty thousand items is a collection of forty thousand rights questions, most of which have no obtainable answer.

The instinct is item-level clearance, and it is the decision that ends programmes. At collection scale it cannot be completed, and an institution premised on it publishes nothing, serves nobody, and carries the same exposure it started with plus years of storage.

The framework that works runs in a different order, and each step removes material from the next.

The public domain determination clears everything that needs no further analysis, and it clears more than institutions expect.

The deed of gift audit identifies where the institution already holds rights.

The accessibility programme covers a use with two independent legal bases and essentially no exposure.

The tiered access model publishes the remainder at a level the assessment supports.

And the takedown operation handles what the analysis missed, which is the cheapest and most effective control in the entire programme.

By the time the residual is reached — in-copyright, untraceable, culturally unrestricted, commercially unexploited — it is a fraction of the original and a manageable one.

The Public Domain Sweep

Do this before anything else, and budget it as a project rather than expecting it as a by-product.

Expired outright. Requires a publication date only. Clears the oldest material immediately.

Published without notice. For much of the twentieth century, publication without a copyright notice injected a work into the public domain. Photographs, postcards, trade catalogues, and ephemera are heavily represented, and the determination is visual — examine the object, including versos, mounts, and packaging.

Published and not renewed. Under 17 U.S.C. § 304, works in the renewal window required renewal, most were not renewed, and the records are searchable. This is the highest-yield investigation available in the sector and it is real work.

Federal government works, not subject to copyright, noting the exclusion covers federal employees within the scope of employment and not contractors, states, or municipalities.

Unpublished works under 17 U.S.C. § 303, with life-plus-seventy for known authors and fixed terms from creation for anonymous, pseudonymous, and work-made-for-hire material — which matters enormously for archives, because an unpublished photograph by an unknown photographer from a known era has a calculable expiry that requires finding nobody.

The foreign trap. Golan v. Holder upheld restoration of copyright in foreign works that had entered the public domain through formalities defects. Restoration was automatic and unregistered, so a foreign work apparently cleared by a no-notice determination may not be. Segregate foreign-origin material and analyse it separately.

Eldred v. Ashcroft upheld term extension, which moved the boundary without changing the method.

Record each determination with its basis and date, at item or collection level, so the work is done once and is auditable.

And re-run the sweep annually, because material enters the public domain every year without anyone doing anything.

The Deed of Gift Audit

The most consequential document in the programme and the one least often read before it starts.

Three positions, and every collection is in one of them.

Copyright assigned. Confirm by reading the operative language against 17 U.S.C. § 204, which requires a signed writing for a transfer of copyright ownership. An accession receipt is not one.

Licence granted. Record the scope — research use, exhibition, non-commercial purposes, term, territory — because it sets the ceiling on the access tier.

Nothing granted. Physical custody alone, with copyright remaining with the donor, the creator, or their heirs.

The donor is frequently not the owner. A family donating a relative's papers holds the objects. Copyright in letters written to the relative belongs to their writers; copyright in photographs of the family belongs to the photographers. Which makes personal papers the hardest category, with dozens of separate owners in a single box.

Restrictions bind independently of copyright. Embargo periods, access limitations, use restrictions, attribution requirements, and occasionally conditions about characterisation. These are contractual, and an institution publishing material under a live donor embargo has a problem no fair use analysis reaches.

Sample legacy deeds by collection and categorise, rather than attempting to read all of them before starting.

And fix the template prospectively. Assign copyright where the donor holds it, take a broad licence where they do not, address digitisation and online access expressly, state what happens to material the donor did not own, and refuse restrictions the institution cannot operationally honour.

Section 108, Fair Use, and What Each Actually Gives

17 U.S.C. § 108 is narrower than its reputation.

What it permits. Preservation reproduction of unpublished works in the collection, in up to three copies. Replacement reproduction of published works that are damaged, deteriorating, lost, stolen, or in an obsolete format, after a reasonable effort to obtain an unused replacement at a fair price. Reproduction of articles and small excerpts for users, subject to conditions.

What it does not permit. General digitisation for online access. Digital copies made under the preservation and replacement provisions may not be made available to the public outside the premises of the library or archive.

Which is the whole problem. The on-premises limitation was written when digital preservation and digital access were different activities. Section 108 supports the scanning and not the publishing.

Section 108(h) is real and underused, permitting reproduction, distribution, display, and performance of published works in the last twenty years of term for preservation, scholarship, or research, where the work is not subject to normal commercial exploitation, a copy cannot be obtained at a reasonable price, and no rightsholder has given notice.

And section 108 expressly preserves fair use, so the correct posture is section 108 for preservation copying and 17 U.S.C. § 107 for access, assessed independently.

Authors Guild v. HathiTrust is the sector's most important decision. Full-text search across a digitised corpus of library holdings: fair use. Accessible copies for print-disabled readers: fair use. Both holdings sit at the centre of what a modern library does.

Authors Guild v. Google held scanning and snippet display for search fair use, emphasising that snippets did not substitute for the works.

Andy Warhol Foundation v. Goldsmith narrowed transformative use by directing attention to shared purpose and substitution, which requires honesty in the sector's assessments: displaying an archival photograph so people can look at it serves the photograph's original purpose, and that is a weaker position than search.

Campbell v. Acuff-Rose Music supplies the framework.

Write one assessment per collection, working all four factors, stating exclusions expressly, approved and dated by whoever bears the risk — because 17 U.S.C. § 504(c)(2) remits statutory damages entirely where a nonprofit library, archive, or educational institution employee acting within the scope of employment had reasonable grounds to believe the use was fair, and the written assessment is what supplies those grounds.

The Diligent Search That Buys You Nothing

There is no United States orphan works legislation. No safe harbour, no statutory search standard, no register of completed searches. A diligent search confers no permission and creates no defence.

Three reasons to do it anyway, each independently sufficient.

It finds owners, and a meaningful fraction of apparent orphans are traceable with ordinary effort. Permission beats every exception.

It informs the fourth factor. Unavailability from any authorised source is directly relevant to market harm, and a documented failed search is evidence.

It bears on remedies, supplying the reasonable grounds that 17 U.S.C. § 504(c)(2) requires.

Set a tiered standard in writing. Tier one — institutional records, finding aids, the object and its markings, Copyright Office online records, a general web search on any name present — applied to everything. Tier two — renewal records, collective licensing databases, publisher and successor research, trade directories, contact with known heirs and professional associations — before open online access. Tier three — genealogical and probate research, correspondence with related repositories, direct outreach, published notice of intent — for commercially valuable material and anything to be licensed out.

Use documented sampling for homogeneous collections. Search a defined sample, record the method and yield, extrapolate, and search item-level only where the sample flags an identifiable creator.

Record sources, dates, terms, results, and reasoning — the record of what was done, not the assertion that a search was diligent.

And set a re-search interval, because records improve and a search that failed in 2016 might succeed now.

Accessibility, the Strongest Position Available

17 U.S.C. § 121 permits an authorised entity — a nonprofit or governmental body whose primary mission is providing specialised services relating to training, education, or adaptive reading or information access needs of persons with disabilities — to reproduce and distribute previously published literary works in accessible formats exclusively for eligible persons. No permission, no licence, no search.

17 U.S.C. § 121A implements the Marrakesh framework and permits cross-border exchange with authorised entities abroad.

And Authors Guild v. HathiTrust held the same activity fair use independently, which means the programme rests on a statutory exception and a Second Circuit holding at once.

Two independent bases is as good as a copyright position gets, and this is consistently the last programme institutions fund.

Produce genuinely usable formats — structured text, described images, tagged documents — rather than image-only scans, and apply the statutory notices the section requires.

Where conversion requires defeating a technical measure, 17 U.S.C. § 1201 applies separately, with accessibility exemptions granted through the triennial process that expire on a three-year cycle.

Access Tiers and Rights Metadata

Publication is not binary, and a tiered model lets material become findable long before it becomes freely reproducible.

Tier zero — metadata only. Catalogue records, finding aids, descriptions. Discoverable, citable, requestable, with zero copyright exposure because descriptions are not the work. This should cover the entire collection and should happen first.

Tier one — on-premises access. Full digital copies on reading room terminals. Supports section 108's framing and eliminates distribution.

Tier two — authenticated access. Full copies to verified researchers or affiliated users, logged, with terms accepted at the point of access.

Tier three — limited public access. Thumbnails, reduced resolution, snippets, or excerpts publicly available with full resolution on request. This is the Authors Guild v. Google posture and it carries most of the material.

Tier four — open access. Full resolution, downloadable, openly licensed or marked public domain.

Assign the tier per collection with recorded reasoning, and design promotion as a metadata edit rather than a re-digitisation project.

Store the rights record in structured fields in the collections management system: status, basis, determination date and determiner, search tier and date, access tier and reason, deed restrictions, claims and resolutions, and review date.

Use the standardised rights statement vocabulary, which is interoperable, aggregator-compatible, and honest about uncertainty in a way free-text notes are not.

And do not overclaim. Bridgeman Art Library v. Corel held that faithful photographic reproductions of public domain works lack the originality Feist Publications v. Rural Telephone Service requires. A blanket copyright notice across a public domain collection is unenforceable and corrosive to institutional credibility.

Takedown, Consultation, and Governance

The takedown operation is the single most effective risk control in the programme and it costs almost nothing.

Publish the policy prominently on every item page and collection landing page, in plain language. Staff a monitored address with a named owner and a service level in hours. Restrict access first and analyse second, because moving an item to metadata-only costs nothing. Verify the claimant's chain of title, because a substantial share of claims come from people who assume they hold rights. Offer the full range of resolutions — attribution, licence, reduced tier, permanent removal — since attribution alone resolves a great many. And log everything, because the claim rate is the empirical answer to the board's risk question.

Distinguish the voluntary policy from 17 U.S.C. § 512, which governs third-party uploads and requires a designated agent, a repeat infringer policy, and expeditious removal on notice. Both should exist and they are different documents.

Run cultural sensitivity review alongside the rights review, not after it, because it can change what gets digitised at all.

Indigenous cultural materials may carry community protocols governing who may view, reproduce, or discuss them, and those protocols do not map onto copyright and are not extinguished by it. NAGPRA addresses human remains, funerary objects, sacred objects, and objects of cultural patrimony, and its consultation architecture is the model even outside its direct scope.

Records of institutional harm — residential schools, asylums, hospitals, carceral and child welfare systems — contain deeply personal information about identifiable people who did not consent.

Consult before digitisation, accept that the answer may be no, build the technical capacity for community-set restrictions before promising them, support Traditional Knowledge labels, and record outcomes in the same rights metadata.

Take a policy to the board, not a project, with explicit escalation triggers for counsel review and one named accountable owner.

The Institution's Own Rights

Institutions spend so long on rights they do not hold that they neglect the ones they do.

Register the institution's original outputs — catalogue essays, original finding aid content, commissioned photography, exhibition materials, publications — because 17 U.S.C. § 411 makes a completed registration a precondition to suit and Fourth Estate Public Benefit v. Wall-Street.com confirmed that means registration rather than application.

Close the contractor gap. Community for Creative Non-Violence v. Reid means commissioned photographers, designers, and writers own their contributions absent a written assignment satisfying 17 U.S.C. § 204.

The institution's name and marks are trademarks for museum, library, and educational services, are frequently licensed for merchandise and exhibitions, and are frequently unregistered.

Image licensing run as a commercial function should be kept clearly separate from the open access programme, so the public offer and the commercial offer do not contradict each other — and should not assert copyright in faithful reproductions of public domain works.

Exhibition and loan agreements carry image rights, catalogue rights, and reproduction permissions that outlive the exhibition, and they are frequently signed by curatorial staff without review.

And funder conditions bind independently. Grant terms requiring open access, specific licence terms, or repository deposit are negotiable at application and immovable at reporting.

Format-Specific Handling

Some categories behave badly enough to warrant their own assessment, and folding them into a general workflow is how a programme acquires its worst surprises.

Audiovisual material. A single film carries copyright in the screenplay, the score, the sound recording of the score, the cinematographic work, and any licensed-in footage or artwork. Music is the usual blocker, because synchronisation and master rights are separately owned and actively administered. A silent print of an orphan film is a very different proposition from the same film with its soundtrack, and splitting the deliverable is often the route through.

Sound recordings. Pre-1972 recordings were governed by state law until federal treatment brought them into a scheme with staged expiry, and the analysis genuinely turns on the recording date. Oral histories carry the interviewee's rights alongside the interviewer's, and a release signed in 1974 for scholarly use is not obviously a release for global streaming.

Correspondence. Letters are unpublished literary works owned by their writers, which means an archive of a public figure's incoming mail is an archive of other people's copyrights. The finding aid can describe them; publishing them is a per-author question.

Newspapers and periodicals. Freelance contributions may never have transferred, and the periodical's own renewal status does not determine the status of the individual pieces inside it. Advertisements are separately owned again.

Theses and dissertations. Owned by their authors, governed by deposit agreements that vary enormously by era and are frequently narrower than the institution assumes.

Born-digital accessions. Email, drives, and social media exports arrive with personal data, third-party material, and credentials mixed in. These need a privacy review before a rights review, and appraisal and disposal are legitimate answers.

Photographs of artworks. Bridgeman Art Library v. Corel governs the photograph; the underlying artwork has its own term, and a recent photograph of a mid-century painting does not make the painting available.

Maps, plans, and technical drawings. Frequently government-produced and public domain, and frequently misfiled because a commercial publisher issued the edition.

Ephemera. Posters, tickets, menus, packaging, flyers. Highest yield for the no-notice determination and lowest yield for owner identification. Assess as a class rather than item by item.

Three-dimensional objects. A museum object may embody a sculptural work, a useful article, an architectural work, or none of these. Photographs of a public domain object are the institution's own works only to the extent they have original expression, and Bridgeman governs the faithful reproduction of a two-dimensional work in a way it does not obviously govern a photograph of a sculpture.

Common Errors

Confusing custody with copyright. The institution owns the object. Somebody else owns the work.

Attempting item-level clearance. It cannot be completed at collection scale, and a programme premised on it is a decision not to proceed dressed as diligence.

Skipping the public domain sweep. It is the cheapest work available and it clears more than institutions expect, particularly the no-notice and non-renewal categories.

Reading 17 U.S.C. § 108 as a publication authority. It permits preservation and replacement copying and expressly bars making those digital copies available off-premises. The access position rests on 17 U.S.C. § 107.

Treating the diligent search as a permission mechanism. There is no safe harbour. The search finds owners, supports the fourth factor, and evidences the reasonable grounds that 17 U.S.C. § 504(c)(2) requires.

Overclaiming on reproductions. Asserting copyright in faithful reproductions of public domain works contradicts Bridgeman, is unenforceable, and damages credibility with exactly the audience the institution serves.

Holding the finding aid back. Descriptions are not the work, tier zero carries no copyright exposure, and findability is most of the value.

Leaving the takedown address unstaffed. The cheapest and most effective control in the programme, and an unanswered rightsholder message is the most reliable way to convert a correction into a claim.

Sequencing consultation after publication. It presents a community with a completed fact and forecloses the outcome consultation exists to permit.

And funding accessibility last. It has a statutory exception, an independent fair use holding, the least exposure of anything the institution does, and the readers with the greatest need.

By Institution Type

Research libraries. The HathiTrust holdings sit at the centre of what these institutions do, and the search and accessibility positions are strong. The recurring failures are section 108 misread as a publication authority and an accessibility programme funded last.

Public library systems. Local history collections, photograph archives, and community material, frequently with no deeds at all. The accessible services function almost certainly qualifies as an authorised entity under 17 U.S.C. § 121 and frequently does not realise it.

University archives. Institutional records, faculty papers, theses, and student material, with the added complication that the institution is both custodian and employer, and its own work made for hire position on faculty output is unsettled.

Art museums. The hardest fair use position, because displaying an artwork serves the artwork's original purpose, and Andy Warhol Foundation v. Goldsmith makes that a weak transformative claim. The strength lies in the public domain sweep, the deed audit, and honest tiering. Image licensing revenue makes the Bridgeman overclaiming problem particularly acute here.

History museums and historical societies. Mixed collections spanning every format in the section above, small budgets, and the largest gap between what the collection could serve and what the institution can process. Sampling and collection-level assessment matter most here.

Government archives. Substantial holdings of federal government works that are not subject to copyright at all, alongside donated material that is. Separating the two is the single highest-yield exercise available.

Indigenous and community archives. Frequently holding material about their own communities under protocols the mainstream framework does not contemplate, and increasingly the counterparties in consultations rather than the subjects of them.

Corporate and institutional archives. Records the organisation authored and owns outright, plus incoming correspondence and third-party material it does not, plus personal data with retention obligations the collection has never applied.

Digital repositories and aggregators. Receiving material from many contributors under many rights positions, which makes the standardised rights statement vocabulary essential rather than merely useful, and which makes the contributor agreement the operative document.

Budget, Staffing, and What Actually Costs Money

Digitisation budgets get built around scanning throughput, which is the one line that behaves predictably and rarely the constraint.

Rights work is the constraint, and it is staff time rather than equipment. Budget it explicitly rather than assuming a cataloguer absorbs it.

The renewal research line is a defined project with a defined deliverable and the highest return per hour in the programme, because every non-renewal it finds converts directly into publishable material. It is also the line most often cut because it looks like overhead.

The metadata line. Determinations recorded in a spreadsheet on a shared drive are lost within one system migration. Recording them in the collections management system in structured fields costs more up front and is the difference between doing the work once and repeatedly.

The consultation line. Community consultation takes months, involves travel and relationship-building, and cannot be compressed by adding staff. Where a collection needs it, the schedule must reflect that from the start.

The takedown line. A monitored address and a named responder. Trivially cheap and the highest-leverage spend in the programme.

What does not need a line. Blanket permission campaigns for material with no traceable owner. A mailing where most of it returns undeliverable is postage, not diligent search.

Staff training matters more than policy documents. The people making item-level calls are cataloguers and digitisation technicians, not counsel, and they need a decision tree they can apply, a clear escalation route, and explicit permission to hold something back without justifying it to a committee.

Give them three questions. Is there a name on it. Is there a date on it. Does anything about it feel like it should not be public. Anything answering yes to the third goes to the named owner regardless of the rights position.

Cadence

Annually, six items.

Re-run the expiry sweep, because material enters the public domain every year without anyone doing anything.

Re-search the flagged high-value orphans, because records improve.

Revisit consultation outcomes, because community positions change and consent given once is not consent forever.

Read the claim log and adjust the tiers on evidence: collections generating claims move down, collections generating requests and no claims move up.

Report five numbers. Items digitised; items published by tier; public domain determinations made; searches completed and owners found; claims received and resolved.

And confirm the standing obligations have named successors — the takedown address, the consultation relationships, the annual sweep, and the review diary. Each outlives the person who set it up, and an obligation with no successor lapses silently.

Expect the claim number to be small. Institutions running these programmes for years typically report claims in the single digits against tens of thousands of items, resolved overwhelmingly by attribution or removal. That is not a promise; it is the empirical base rate against which the anxiety should be calibrated.

A Closing Note

The material in these collections was preserved so that people could reach it.

The framework in this toolkit is not a set of obstacles to that. It is the sequence that gets the most material to the most people with the least exposure, in an order that removes work rather than adding it.

And the institution that spends three years attempting item-level clearance publishes nothing, serves nobody, and carries exactly the exposure it started with — plus three years of storage costs on material nobody can see.

Do the sweep, read the deeds, fund accessibility, tier the rest, and staff the takedown address. That is the whole programme, and it is a defensible institutional posture rather than a compromise.

The Objections, Answered

"We should wait for orphan works legislation." Proposals have circulated for two decades without enactment while the material continues to deteriorate. Waiting is a decision with its own cost.

"Our insurer will not cover this." Ask them, then decide with the answer. Some institutional policies respond to copyright claims and many do not, and the standard advertising injury grant is narrower than institutions assume. Either way, the exposure is usually smaller than the premium conversation implies.

"A rightsholder could sue for statutory damages on every item." In theory. In practice 17 U.S.C. § 504(c)(2) remits statutory damages for a qualifying institution acting on reasonable grounds, the written assessment is the record of those grounds, and a claimant would have to establish registration and ownership item by item.

"We are not a library, so section 108 does not apply." Probably right and largely irrelevant, because the access position rests on 17 U.S.C. § 107 rather than on section 108 anyway.

"The donor's family will object." Consult them. A family that is asked usually cooperates; a family that discovers the publication is a different conversation.

"We will lose licensing revenue." Check the actual figure. For most collections it is smaller than the cost of administering the licensing, and open access frequently increases requests for the high-value material that does generate revenue.

"Our peers are not doing this." Many are, and have been for years, and their annual reports carry the claim numbers. Ask them directly rather than inferring caution from silence.

"What if we get it wrong on one item." Then a rightsholder writes to the published address, the item goes to metadata-only within the day, and the institution resolves it. That is what the takedown operation is for, and treating one predictable correction as a programme-level failure is how nothing gets published.

"We do not have the staff." Then run the sweep, publish tier zero across the whole collection, and staff the takedown address. Those three cost less than a single conservation treatment and they change what the institution can offer more than anything else on the list.

Sequencing, and Why the Order Matters

The order in this toolkit is not arbitrary. Each step reduces the volume the next has to handle.

The public domain sweep removes the material that needs nothing further, and for twentieth-century photographs and ephemera it removes a great deal.

The deed review removes the material where the institution already holds rights, which for institutionally generated records and well-papered accessions is another substantial slice.

The accessibility programme covers a use with two independent legal bases and no meaningful exposure, and it serves the readers with the greatest need first.

Tier zero publication makes the entire collection findable at no copyright cost, which is most of the value most users are looking for.

The tiering places the remainder at a level the written assessment supports.

And the takedown operation handles the residue of error, which in practice is very small.

By the time the hard category is reached — in-copyright, untraceable, culturally unrestricted, commercially unexploited, and intended for open access — it is a fraction of the original collection.

An institution that inverts this order starts with the hardest category, exhausts its budget on it, and never reaches the material it could have published on day one.

Which is the most common failure in the sector, and it is not a failure of doctrine or of resources. It is a sequencing failure, made once, at the point the programme was designed.

Design it in this order and the same budget produces several times the published output, with a smaller residual risk and a defensible record of how each decision was reached.

Working With Aggregators and Partners

Most institutions now contribute to something larger — a state or regional aggregator, a national portal, a subject repository, or a commercial digitisation partner — and each relationship carries terms worth reading.

Aggregator contributor agreements typically require standardised rights statements, a warranty about the contributor's right to supply the material, and a licence permitting the aggregator to display and to redistribute to downstream partners. The warranty is the provision that matters: an institution warranting it holds all necessary rights, across a collection it assessed at collection level with documented sampling, is warranting more than it can support. Negotiate it down to a representation that the institution has applied a documented rights assessment process.

Commercial digitisation partnerships — where a vendor digitises a collection in exchange for a period of exclusive commercial access — carry three recurring problems. The exclusivity period may exceed what a funder's open access condition permits. The vendor may claim rights in the digital surrogates, which for faithful reproductions of public domain works contradicts Bridgeman. And the institution may not receive files in a format it can use independently at the end, which is the exit and portability problem in a collections setting.

Reciprocal loan and exhibition arrangements carry image and catalogue rights that outlive the exhibition and are frequently signed without review.

Consortial licensing for published content is governed by the subscription terms, and those terms should carry an accessible-format conversion carve-out and should permit the linking that makes course and research use possible without copying.

And joint digitisation projects between institutions need the same clarity a coalition needs: who owns the surrogates, who holds the metadata, whose rights statements govern, and what happens if one partner withdraws.

Read every one of those agreements against the institution's own rights position before signing, because each of them asks the institution to warrant something, and the warranty is invariably broader than the assessment behind it.

A representation about process is defensible; a warranty about title across forty thousand items is not, and no institution in the sector could give one honestly.

Say so plainly in the negotiation, and most counterparties accept the amendment without argument.

A Suggested Reading Path

Start with the doctrine in The Rights You Cannot Trace.

Then the programme in Running a Digitisation and Access Programme.

Then the audit in the cultural heritage digitisation checklist.

For the public domain work, Is It in the Public Domain and the public domain clearance checklist.

For the fair use analysis, Fair Use After Warhol and the Fair Use and Permissions Toolkit.

For the accessibility programme, Making Works Reachable and the Accessibility and Inclusive Content Toolkit.

For duration mechanics, the Copyright Duration and Public Domain Toolkit.

And for the institution's own assets, the Copyright Fundamentals Toolkit.

Primary Authorities

| Authority | Proposition | |---|---| | 17 U.S.C. § 101 | Definitions; work made for hire | | 17 U.S.C. § 102 | Subject matter | | 17 U.S.C. § 106 | Exclusive rights | | 17 U.S.C. § 107 | Fair use | | 17 U.S.C. § 108 | Libraries and archives; on-premises limit | | 17 U.S.C. § 109 | First sale; display | | 17 U.S.C. § 110 | Performances and displays | | 17 U.S.C. § 121 | Accessible formats | | 17 U.S.C. § 121A | Cross-border accessible copies | | 17 U.S.C. § 201 | Ownership; transfers | | 17 U.S.C. § 204 | Signed writing for transfers | | 17 U.S.C. § 302 | Duration | | 17 U.S.C. § 303 | Unpublished works | | 17 U.S.C. § 304 | Renewal terms | | 17 U.S.C. § 411 | Registration precondition to suit | | 17 U.S.C. § 504 | Damages; nonprofit remission | | 17 U.S.C. § 512 | Safe harbour and takedown | | 17 U.S.C. § 1201 | Circumvention; exemptions | | Authors Guild v. HathiTrust | Search and accessibility as fair use | | Authors Guild v. Google | Snippet display for search | | Andy Warhol Foundation v. Goldsmith | Transformative purpose narrowed | | Campbell v. Acuff-Rose Music | Transformative use framework | | Bridgeman Art Library v. Corel | Faithful reproductions lack originality | | Feist Publications v. Rural Telephone Service | Originality requirement | | Golan v. Holder | Restoration of foreign works | | Eldred v. Ashcroft | Term extension upheld | | Fourth Estate Public Benefit v. Wall-Street.com | Registration means registration | | Community for Creative Non-Violence v. Reid | Contractors own their contributions | | NAGPRA | Repatriation and consultation | | Standardised rights statements | Interoperable rights metadata | | Traditional Knowledge labels | Community protocols | | Orphan works diligent search proposals | The unenacted framework |

Forms and Templates

The License Agreement Template supplies the structure for licensing collection images out and for licensing material in, with the scope, term, and attribution provisions an institution needs on both sides. A revised deed of gift is the highest-value template in the sector: it should assign copyright where the donor holds it, take a broad licence where they do not, address digitisation and online access expressly, state what happens to material the donor did not own, and refuse restrictions the collections system cannot express. Exhibition and loan agreements should be treated as a separate template with image, catalogue, and reproduction rights that survive the exhibition addressed expressly, and commission agreements for photographers, designers, and writers should carry assignments satisfying 17 U.S.C. § 204.

Related Toolkits and Checklists

The Copyright Fundamentals Toolkit carries the ownership, registration, and duration doctrine. The Copyright Duration and Public Domain Toolkit supports the sweep that should run first. The Fair Use and Permissions Toolkit covers the assessment that supports the access programme, and the Accessibility and Inclusive Content Toolkit covers the programme with two independent legal bases that should be funded first. For the institution's own outputs and the contractor gap, use the chain of title checklist.