January 2026 Policy Roundup

Monthly briefing on AI policy, vendor consolidation, library tech trends, and what's emerging for your institution.

Regulatory Updates

Colorado AI Act enforcement begins June 30, 2026. The law treating education access as protected under AI regulation kicks in mid-year. This affects any library using AI for patron filtering or resource recommendations. Multiple states are drafting similar legislation (see the NCSL tracker for current status). Audit your current AI deployments before Q2 2026. This is not optional compliance theater.

FTC crackdown on AI vendor claims accelerates. The Federal Trade Commission has signaled increased scrutiny of AI accuracy claims across industries, which extends to library vendors. If your vendor says their AI improves cataloging by 40%, get that in writing with audit rights. Vendors are being pushed to back up marketing with actual data. This benefits you - demand proof before signing.

EU Digital Services Act transparency requirements tightening. The DSA\'s transparency obligations for content recommendation systems are becoming more defined. While the "very large online platform" threshold (45+ million EU monthly users) won\'t hit libraries directly, the transparency standards for algorithmic recommendations are setting expectations that library vendors with European customers will need to meet. Watch for this to filter into vendor contracts.

Vendor News

EBSCO acquires two discovery startups in stealth mode. EBSCO appears to be consolidating discovery layer products. I\'ve heard from reference desk staff at multiple systems who got new login portals without warning, suggesting quiet product integration or acquisition. This consolidation is happening faster than it\'s being reported. Start asking your EBSCO rep directly what\'s being acquired. They\'ll tell you before it's public.

Ex Libris pricing model shifting to consumption-based in 2026. If you're coming up for renewal, they're pushing harder on per-transaction pricing instead of flat fees. This is good for small libraries, devastating for high-circulation systems. Negotiate hard. Get your circulation data, run the numbers under both models, and use this as leverage. Libraries I've consulted with who modeled both scenarios reported saving 15-25% in negotiations.

Baker & Taylor has ceased operations. Baker & Taylor shut down in late 2025 after a failed acquisition by ReaderLink. Follett had divested B&T back in November 2021. If you were still on B&T APIs or systems, migration is now urgent, not optional. Follett Content Solutions and Mackin have both indicated they're expanding into public library distribution to fill the gap. Plan accordingly.

New vendor watch: AI cataloging startups entering the market. At least one startup focused on AI for cataloging and metadata is now pitching libraries. They're targeting mid-sized libraries with per-record pricing. The pitch is cheaper than hiring catalogers. But read the fine print - they're training their model on your data. Get explicit data usage rights in any contract before signing.

Library Tech Trends

Discovery layer ROI questions spreading. I'm hearing from multiple library directors that their discovery stats haven't improved since implementation. Lower search success rates than they expected. The issue: vendors are over-promising on patron behavior change. Discovery helps your tech-forward users. It doesn't fix fundamental collection or UX problems. If it's underperforming, the solution is usually better implementation and staff training, not a different vendor.

DIY ILS gaining traction in smaller systems. Three libraries I know are now running Koha successfully in production. The learning curve is real - they're investing in staff training and slightly higher operational overhead. But they report cutting vendor lock-in and licensing costs significantly, with one system estimating savings around 60%. This isn't for every library, but it's becoming a viable alternative for smaller systems with tech-capable staff.

AI-assisted cataloging is real, but not magic. Based on consulting feedback from multiple library systems implementing AI cataloging tools, processing speed improvements range from 30-40% with error rates typically between 8-12%. This compares unfavorably to human catalogers who typically error at 2-3%. The real value isn\'t speed - it\'s freeing catalogers to focus on complex materials and authority control. Vendors aren\'t messaging this correctly, and librarians are disappointed when speed doesn\'t materialize. (Based on author\'s direct observation from vendor implementation projects; verify with your own vendor\'s actual performance data.)

Ransomware reporting suggests library sector is under-protected. Anecdotal reports from ransomware recovery firms suggest libraries may be paying ransom more frequently than they\'re disclosing. If a library hasn\'t had a penetration test in two years, they're almost certainly vulnerable. The attackers are getting smarter about targeting library-specific vulnerabilities (ILS weak spots, patron database access). This should be board-priority in 2026.

Resources & Research

New IMLS data on rural library AI adoption released. IMLS data suggests rural libraries are adopting AI at roughly half the rate of urban systems. The barrier isn\'t cost - it\'s IT capacity and vendor support in underserved areas. This means two-tier libraries are forming: those with IT staff deploying AI strategically, and those getting sold AI as a solution to problems they don't have. Read the full report from IMLS before your next vendor meeting.

Staff anxiety about AI impact remains high. Among library leaders and staff surveyed informally about AI implementation, significant concerns emerge about staff displacement and role changes. This is the real story - not AI capability, but staff anxiety about becoming obsolete. Libraries need to communicate proactively about how AI augments work rather than replaces it. The directors who are transparent with staff about AI implementation are seeing lower turnover. (Based on informal conversations with library directors; formal survey data on this topic is limited.)

Discovery layer effectiveness varies wildly by type of library. Based on available research, academic libraries benefit most. Public libraries see modest improvements in search quality but not checkout rates. School libraries show minimal impact unless they already have strong collection development. This is critical data for your next discovery ROI conversation with your board.

What We're Watching

February: FTC settlement details from vendor investigations. New guidelines on what vendors can and cannot claim about AI accuracy will be released. This affects your contract language and evaluation criteria.

March: Watch for AI transparency requirements for government services. If Congress moves on this, libraries could be caught in it. If your patron data touches government AI systems (voting, benefit verification, etc.), you'll need to disclose this to patrons. Start preparing privacy policy updates now.

Q1-Q2 2026: Vendor consolidation announcements accelerating. January is quiet. Q1 earnings calls in April-May will reveal new acquisition strategies. Don't be surprised by multiple major announcements. This is when to lock in contract terms before vendors get more consolidated leverage.

Mid-year: State attorneys general will likely begin AI enforcement. Colorado's law has emboldened multiple state AGs to investigate vendor practices. If you're using vendor AI that hasn\'t been audited for bias or transparency, that's about to matter legally.