Causaly bought the right to read

❝

In THIS ISSUE
Signal of the Week, Causaly licenses full text of 400+ journals for its AI agents.
Framework of the Week, the Evidence Supply Chain.
5 Things Worth Reading, deals, regulators, and the MSL shift.
Tool of the Week, Causaly. Reading time: ~8 min read

🎯 Signal of the Week

Causaly did not buy a better model this week. It bought the right to read.

Welcome to Issue 04,

For two years the AI conversation in pharma has been about models: which foundation model, how many parameters, whose agent framework. The signal this week says the fight has moved somewhere less glamorous and far more decisive. Causaly, one of the more serious biomedical evidence platforms, signed a licensing deal with Sage that lets its AI agents read the full text of more than 400 Sage and Mary Ann Liebert life-sciences journals, not just the abstracts. It is Causaly's first tie-up with a major scholarly publisher, and it is a template, not a one-off.

Why this matters for Medical Affairs specifically: literature monitoring is the single highest-leverage AI use case for MSLs and medical information teams, and until now most tools have been reasoning over abstracts, titles, and whatever open-access fragments they could legally touch. An abstract tells you a study exists. The methods section, the results tables, and the supplementary data tell you whether the study is any good. An agent that can only read abstracts is a very confident intern who never opened the paper. An agent with licensed full text can weigh methodology, flag an underpowered subgroup, and surface the table that actually changes your scientific narrative.

Here is the opinionated part. The moat in enterprise AI for our function is not the model, which is rapidly commoditizing, and it is not the agent orchestration, which everyone is copying. The moat is licensed, machine-readable access to the evidence base, plus a clean audit trail of what the system was allowed to read. Causaly buying reading rights is the first visible move in a land grab for content licensing.

_{Figure 1. The model commoditizes; the durable advantage sits in licensed content plus an audit trail. Deal facts from the Causaly and Sage announcement,}_{reported by GEN}_.

Expect your vendors to start competing on which journals they can legally ingest, and expect procurement and compliance to start asking a question they have never asked before: what exactly is our AI allowed to read, and can we prove it. If you run Medical Affairs, put a content-rights audit on this quarter's list. The teams that treat full-text licensing as an afterthought will ship confident, ungrounded summaries, and they will be the ones explaining themselves to compliance later.

🧠 Framework of the Week

The Evidence Supply Chain

Every Medical Affairs team is now under pressure to put AI into literature monitoring, medical information, and evidence synthesis, and most are judging the tools on the wrong axis: how fluent the answer sounds. Fluency is the easiest thing for a model to fake and the weakest predictor of whether a claim survives a compliance review or a KOL who actually read the paper. The question that matters is not how good the output looks, it is how the output was made.

This week's signal is really a statement about that pipeline. When a publisher licenses full text to an AI platform, the value moves from the model to the evidence it can legally read and the trail it leaves behind. So the framework this week grades the pipeline, not the prose. I call it the Evidence Supply Chain: five links, in order, that every answer travels before it earns your trust. Each link is a place the chain can break, and the first place a regulator or a medical director will look when something goes wrong.

Read it as a diagnostic, not a maturity model. You are not scoring how advanced your AI is; you are scoring whether one specific output is safe to use today. Walk any workflow through the five links and mark each one red, amber, or green before you rely on it.

Five links decide whether a Medical Affairs AI workflow is defensible. Score each one red, amber, or green before you trust an output.

Rights: does the system have a legal, documented license to the sources it reads (full text, not scraped abstracts)? No rights, no trust.
Retrieval: is the answer grounded in retrieved source passages (RAG), or is the model free-associating from training data? Grounding beats fluency.
Reasoning: can the system show its work, quoting the passage and citing the paper, so a human can check the claim in under a minute?
Review: is there a human-in-the-loop checkpoint before anything reaches an HCP, a label, or a slide, with promotional and non-promotional workflows kept strictly separate?
Record: is every read, prompt, and output logged, so you can reconstruct what happened when an inspector or a KOL asks?

^{Figure 2. Five links from source rights to audit record. Score each one before you trust an output. (Source Dr. Issam Chebouti)}

❝

HOW TO USE IT THIS WEEK

Take one live AI workflow in your team (literature surveillance, medical information triage, or pre-call briefs) and score its five links today. Any red at Rights or Record is a stop sign: fix licensing and logging before you scale the workflow, not after. Amber in the middle is fine to run with a human in the loop

_{Figure 3. Why the Causaly and Sage deal is a template, not a one-off. Sources: Causaly and Sage announcement (Jul 2026); Medical Affairs AI adoption survey.}

❝

The model is the part everyone can buy. What it is allowed to read, and whether you can prove it, is the part that is actually yours.

📚 5 things worth to read

Five paywall-free reads, each summarized so you get the value without the click

01
Sage inks licensing deal with Causaly's AI platform
The deal behind this week's signal. Sage licensed the full text of more than 400 Sage and Mary Ann Liebert journals to Causaly's AI agents, the platform's first tie-up with a major scholarly publisher. The detail to keep: agents can now weigh methods, results tables, and supplementary data, not just abstracts, which is exactly the layer that decides whether a study changes your narrative. For Medical Affairs, it signals that full-text licensing, not model choice, is becoming the contested asset.
_{Read on GEN →}

02
EMA and FDA issue joint AI guidance for medicine development
The two agencies proposed ten shared principles for good AI practice across the medicine lifecycle, the first real transatlantic alignment on AI in drug development. The through-line is familiar and worth internalizing: risk-based validation scaled to context of use, human oversight, and traceable data governance. For Medical Affairs, it is the clearest signal yet of the documentation bar your AI-supported evidence work will be measured against.
_{Read on European Pharmaceutical Review →}

03
AI is reshaping scientific publishing in 2026: biology and medicine are growing fastest
The other side of this week's signal. Global scholarly output is on track to cross 6 million articles, and biomedicine leads the AI-driven surge. The part Medical Affairs cannot skip is integrity: researchers estimate roughly 147,000 hallucinated citations entered the literature in 2025, and a Lancet audit found fabricated references climbing from about 1 in 2,828 papers to 1 in 277, with review articles 57% worse. It pairs exactly with this issue's compliance note: verify every citation an AI hands you, because the evidence base your agents read is getting noisier, not cleaner.
_{Read on Futurity Publishing →}

04
Reimagining the role of medical science liaisons in the age of AI
Candid interviews with people building MSL tooling on what AI actually takes off the plate (background research, admin, drafting) and what stays stubbornly human (trust, peer-to-peer scientific exchange). The honest read is that pharma has been slow here, and pricing pressure may be the thing that forces the pace. A good gut-check on hype versus reality.
_{Read on pharmaphorum →}

05
Regulating AI in drug discovery: what FDA, EMA and ICH guidance means
The cleanest single map of who requires what, and where each regulator still leaves questions open: FDA's framework still in draft and excluding discovery, EMA's reflection paper explicitly including it, ICH leaning on technology-neutral trial principles. The takeaway for evidence teams: build governance and documentation now, because "wait and see" is no longer a viable strategy.
_{Read on Drug Discovery News →}

🔧 TOOL OF THE WEEK Causaly

A practical pick for MSLs and Medical Information teams, chosen with the last read in mind

WHAT IT IS

A biomedical evidence and research platform built on AI agents that read the literature and return cited, source-grounded answers rather than a list of links. With the new Sage license, its agents now reason over full-text methods, results, and supplementary data across 400+ journals, on top of its existing biomedical corpus.

HOW TO USE IT FOR MEDICAL AFFAIRS AND MSLS

Point it at a defined scientific question (a mechanism, a competitor's trial readout, an emerging safety signal) and let it assemble a cited evidence brief before a KOL meeting or an advisory board. Use it for literature surveillance in your therapeutic area, for pressure-testing a scientific narrative against the primary data, and for triaging medical information inquiries. Treat every output as a first draft with citations attached, not a finished answer.

WHERE TO ACCESS IT

causaly.com. Full-text integration ships as an add-on, so confirm which journals your license actually covers before you rely on it.

COMPLIANCE NOTE

This is a retrieval-grounded, licensed-source workflow, which is the safer pattern, but the guardrails are still yours to enforce. Keep it strictly separate from promotional review, require human sign-off before anything reaches an HCP or a document, verify that citations resolve to the real paper, and make sure the read and query logs are retained for audit. Grounded is not the same as governed.

That is Issue 04. If a colleague forwarded this to you, you can get it yourself every Monday at newsletter.ichealth-ai.com/subscribe.
One email a week. No sponsorship. See you next Monday.

iCHealth Pathway is a weekly research note by Dr. Issam Chebouti on AI transformation for Pharma, Medical Affairs, and Healthcare leaders.

iCHealth-AI, ichealth-ai.com, [email protected]

_{Issue 04, Monday, August 24th, 2026.}
_{Parts of this issue were produced with AI assistance and checked before publication.}