Regulatory and AI Round-Up: Week of 27 July 2026

Welcome to our weekly round-up for clinical operations, regulatory affairs, quality and pharmacovigilance teams. Each edition covers what moved, what it changes in your systems and SOPs, and the dates worth putting in the diary.

The short version

• The EU AI Act's general date of application arrived on 2 August 2026. The Article 50 transparency obligations and the governance framework are now live. The high-risk deferral we covered last week does not touch either.
• FDA finalised three cancer clinical trial eligibility guidances on 28 July, covering performance status, laboratory values, and washout periods and concomitant medications.
• The MHRA opened an AI regulatory sandbox inside the NHS on 30 July, putting AI-enabled devices into live clinical settings under regulatory oversight.
• Trial demand looks steady. Every CRO reporting in or around the week posted a net book-to-bill above 1.0.

In this round-up: EU AI Act · Trial eligibility · AI oversight in the UK · Trial demand · Sponsor activity · Research funding · Dates to diarise

1. The AI Act general application date has arrived

Live since 2 August 2026. Article 50 and the governance framework were never part of the deferral.

Last week's edition covered the mechanics of Regulation (EU) 2026/1744 in detail, so this is the short version: the date came, and the thing that was deferred is still the only thing that was deferred.

Chapter III, Sections 1, 2 and 3 now apply from 2 December 2027 for stand-alone Annex III high-risk systems and 2 August 2028 for high-risk AI embedded in Annex I regulated products. Everything else arrived on schedule on 2 August 2026.

For most life sciences organisations the practical question this week is narrow. Where you act as a provider or a deployer of a system that interacts with people or generates or manipulates content, Article 50 may require disclosure, marking or labelling, and that obligation is now in effect. One accommodation applies: providers of generative AI systems already on the market before 2 August 2026 have a four-month transitional period to adapt to the Article 50(2) marking obligations, which ends on 2 December 2026.

If your organisation paused its AI Act work in June on the strength of the deferral headlines, this is the week to confirm that the pause was scoped to the high-risk package and did not quietly extend to disclosure.

Source: Regulation (EU) 2026/1744 on EUR-Lex

2. FDA finalises three cancer trial eligibility guidances

Published 28 July 2026. Three guidances, finalising drafts issued in April 2024.

FDA announced three final guidances in the Federal Register on 28 July 2026, all developed with the Oncology Center of Excellence and all aimed at the same problem: eligibility criteria that exclude patients without a scientific justification for doing so.

The three are:

• Performance Status (Docket FDA-2024-D-1377, 91 FR 47251)
• Laboratory Values (Docket FDA-2024-D-1402)
• Washout Periods and Concomitant Medications (Docket FDA-2024-D-1376)

Each recommends that sponsors justify restrictions rather than carry them forward by convention. Performance status criteria should be broadened where the mechanism and safety profile allow. Laboratory value cut-offs should reflect the actual risk of the investigational product rather than a default range. Washout periods and concomitant medication prohibitions should be based on pharmacology, not on precedent from an earlier protocol.

The context FDA has been putting alongside this work is worth quoting for scale. The Oncology Center of Excellence has noted that fewer than 5 percent of cancer patients currently undergoing treatment are enrolled in a clinical trial, against more than 70 percent who express willingness to participate.

What this changes operationally

This is a protocol design guidance, but the consequences land squarely in trial execution.

Broader criteria change your screening arithmetic. Screen failure rates, the eligibility logic configured in your data capture system, the pre-screening questions in recruitment workflows and the site-facing eligibility checklists all derive from the criteria in the protocol. Loosening them is not a one-line protocol edit.

Broader criteria also change your safety profile. A population with wider performance status and less restrictive laboratory thresholds is a population with more comorbidity and more concomitant medication. Expect more adverse events that require assessment, and check that your case processing and medical monitoring capacity is sized for it.

There is a connection to last week's item on ICH M11 here. As protocol content becomes structured and machine-readable, eligibility criteria are among the first fields that can be carried directly into downstream study build rather than re-entered by hand. Guidance that asks sponsors to revisit those criteria more often makes that automation more valuable, not less.

These are non-binding guidances and comments can be submitted at any time.

Source: Federal Register notice, 28 July 2026

3. The MHRA opens an AI regulatory sandbox inside the NHS

Announced 30 July 2026. Up to 10 manufacturers in the initial phase.

The MHRA, with NHS England in London and the capital's three Health Innovation Networks, launched a regulatory sandbox that will place AI-enabled medical devices into live NHS clinical settings under MHRA oversight, generating real-world safety and effectiveness evidence as they run.

Up to 10 AI medical device manufacturers will be selected for the initial phase, working with NHS providers across London. Expressions of interest are open.

The direction here is more interesting than the announcement. The AI Airlock work the MHRA has been running was about testing how to regulate AI devices. This puts the testing into live clinical use, which means the evidence generated is real-world evidence about products in service rather than simulation.

For anyone building or deploying AI in a UK clinical or health setting, this is a route to generating regulator-visible evidence rather than assembling it retrospectively. It is also a reasonable signal of what the MHRA will expect to see from AI-enabled products outside the sandbox in due course: monitored performance in context, not a validation report from a single point in time.

Also from the MHRA this week: the agency published a review of the risk of dementia associated with bladder anticholinergics on 27 July. Worth a look for anyone whose portfolio or signal detection work touches that class.

Source: GOV.UK, 30 July 2026

4. Trial demand: what the CRO numbers said

Every reporting name posted a net book-to-bill above 1.0.

Three large CROs reported second quarter results during the week, and one more reported the week before. Read together, they are the clearest read we have had in a while on where trial volume is going.

A book-to-bill above 1.0 means new business is being booked faster than backlog is being converted into revenue. Sustained across four companies, that points to a pipeline of trials being commissioned rather than deferred, and several management teams attributed it explicitly to improving biotech funding conditions and a lower cancellation rate.

Two cautions before anyone builds a plan on it. Bookings are a leading indicator and revenue lags them by quarters, not weeks, so this is a statement about 2027 activity more than 2026. And strong bookings did not translate into strong margins everywhere: ICON's revenue grew 1.2 percent while adjusted EBITDA fell 21.7 percent, which is a reminder that winning work and running it profitably are separate problems.

For clinical operations teams the read is straightforward. If your organisation is on the sponsor side, the capacity you are competing for is being contracted now. If you are planning system capacity, headcount or study start-up throughput for next year, plan for volume.

Sources: IQVIA 8-K, ICON results, Fortrea results

5. Sponsor activity worth noting

Four events from the week, each of which creates work for somebody's clinical operations function.

argenx to acquire Forte Biosciences, 27 July. An all-cash deal at $77.00 per share, valuing Forte at roughly $2.2 billion, around an 86 percent premium to the volume weighted average price since positive Phase 1b vitiligo data on 9 July. The asset is FB102, a first-in-class anti-CD122 monoclonal antibody, with Phase 2 celiac and Phase 1b alopecia areata readouts expected in the second half of 2026. Close is expected in the third quarter. Acquisitions of this shape mean two clinical systems estates that have to be reconciled, and a trial master file that has to be inspection ready on both sides of the transition.

TScan doses first patient in Phase 3, 29 July. ALLOHA-2 (NCT07702578) is a pivotal trial of TSC-101 for preventing relapse after allogeneic haematopoietic cell transplant in AML and MDS, with relapse-free survival as the primary endpoint. A first patient dosed in a multi-site cell therapy pivotal is one of the most system-intensive moments in a programme.

Novo Nordisk's ZEUS trial misses its primary endpoint, 31 July. Ziltivekimab lowered IL-6 and hsCRP but did not reduce major adverse cardiovascular events in a trial of more than 6,000 patients followed for up to four years. A non-cash impairment will be taken in the third quarter; two other ziltivekimab trials continue with 2027 readouts. A study of that size does not simply stop. Close-out, data lock and trial master file completeness across a very large site footprint is months of work, and it is the kind of work that gets under-resourced precisely because the programme is no longer a priority.

Replimune's RP1 clears its advisory committee, 30 July. FDA's Cellular, Tissue and Gene Therapies Advisory Committee voted 10 to 3 that the IGNYTE efficacy data for RP1 with nivolumab in anti-PD-1-resistant advanced melanoma are evaluable and clinically meaningful. The vote is non-binding and an FDA decision was expected in early August. Worth tracking if you follow oncologic gene therapy precedent, particularly given the reviewers' disagreement over the stricter response rate analysis.

Sources: BioPharma Dive on argenx and Forte, TScan press release, BioPharma Dive on ZEUS, BioPharma Dive on RP1

Dates to diarise

Two dates came off this table since last week. The EU AI Act general application date of 2 August 2026 has passed, and the QMSR, the UK trials regime and the ICH M11 EU template are all live.

Back next week.

This article is for general informational purposes and does not constitute legal or regulatory advice. Regulatory positions and implementation guidance should be checked against the applicable official sources.