One of the most-talked-about AI stories this week came from Anthropic (Claude) CEO Dario Amodei. In his essay, âWe Must Pace the Frontier,â he argues that AI is advancing faster than safety rules and oversight can keep up.
His point: the companies building the most powerful AI systems can see what is coming before the rest of us can. He wants stronger testing, shared safety standards, independent review, and more coordination between governments.
You can agree with him, disagree with him, or question the timing.
But here is what matters for Medical Affairs: while people debate whether AI development should slow down, the field has already sped up.
Waiting to see what happens is not a strategy.
Welcome to The Frontier Issue.
1. Our Take (The Patrina + Sarah Version)
We'll plant a flag on a few things.
Slowing the frontier is a lab problem. Falling behind is a you problem.
The safety conversation is important. But it does not change what will land on your desk Monday morning.
Whether the major AI companies slow down or not, you still need to know how to use the tools available to you today. Do not let a policy debate become your reason to wait.
The phrase we keep coming back to is expert in the loop, not just human in the loop.
Everyone says, âKeep a human in the loop.â But if that person cannot tell whether the answer is accurate, they are only approving what AI gives them.
In Medical Affairs, the human must be a scientific expert who can:
- Catch a fake or mismatched citation
- Question an oversimplified mechanism
- Spot outdated data or an invented statistic
- Recognize when AI sounds confident but is wrong
AI does not make your scientific judgment less important. It makes it more important. Your expertise is the safety check.
The gap is growing.
We see the same divide in every team we train. Some people have barely opened the approved AI tool. Others use it every day.
That gap is beginning to affect who comes to meetings prepared, who gets heard, and who is asked to join bigger projects.
While AI leaders are thinking about what may happen six months from now, another MSL may already be doing in 20 minutes what still takes you all morning.
Our advice is simple: Get good. Stay the expert. Use AI on purpose.
2. The Field Already Voted
We asked Medical Affairs one question for two years running: How often are you using AI in your work?
Look at what happened in twelve months.
July 2025 (n = 336) â Daily: 55% â Weekly: 32% â Occasionally: 8% â Never: 5%
July 2026 (n = 441) â Daily: 76% â Weekly: 12% â Occasionally: 8% â Never: 3%
Daily use rose 21 percentage points in one year. At the same time, weekly use fell from 32% to 12%, suggesting that many weekly users became daily users.
And only 3% said they never use AI. That means 97% of respondents use AI at least occasionally.
Daily AI use is no longer limited to early adopters. The question is no longer, âShould I use AI?â
The better question is: Can I use it well and still be the smartest person in the loop?
3. Where Is Your Team Today?
Most companies we see fall into one of three groups.
The Locked Vault
AI is still âunder review.â People may not know which tools are approved, how to access them, or what they are allowed to do.
The company may feel protected, but employees may turn to personal accounts or unapproved tools to get the work done. The risk did not disappear. It became harder to see.
The Wild West
Everyone is experimenting, but no one has been trained. Different people use different tools, and there is no shared standard for what is safe, accurate, or appropriate.
There is plenty of energy but also plenty of risk.
Woven In
AI is built into the teamâs real work. The tools are approved. People are trained. Clear standards are in place. And human judgment stays in charge.
AI is not treated as a separate initiative. It helps people do work they already need to do, better and faster.
This is where companies need to go.
Is your team in the Locked Vault, the Wild West, or is AI Woven In?
If you want help moving your team toward Woven In, reply WOVEN, and weâll show you what that can look like.
4. Your Prompt of the Week: Teach AI to Catch Its Own Mistakes
Expert in the loop only works if you actually know where the machine tends to slip. So instead of feeding it your data, have it hand you the cheat sheet. Paste into Mira:
"Act as a skeptical Medical Affairs reviewer whose job is to train me to catch your mistakes. Don't ask me to paste any confidential or company data. Just teach me the specific ways AI tends to get scientific and clinical claims wrong: fake or mismatched citations, outdated data, overstated cause and effect, invented statistics, wrong dosing, and anything else you're prone to. Then give me a short checklist I can save and run on any AI output in under a minute, before I trust it or use it with an HCP."
Two minutes, and you walk away with a reusable checklist instead of a one-time answer. That's what being the expert in the loop actually looks like.
(Quick reminder for everyone: keep confidential, proprietary, and patient information out of any public AI tool. When the data is sensitive, use your company's approved, secure system, not the open web.)
TL;DR: The Frontier Issue
â Amodei's "Pace the Frontier" essay is a real safety debate, and it's a lab problem. It does not change whether you know how to use these tools well.
â Aim for expert in the loop, not just human in the loop. A human who can't evaluate the output is a rubber stamp. Your scientific judgment is the safety layer.
â The field has already voted: daily AI use jumped from 55% to 76% in one year. 97% of Medical Affairs now uses AI. Waiting is not a strategy anymore.
â Which lab are you in? The Locked Vault, the Wild West, or the Woven-In. Only one wins. Reply WOVEN.
The frontier debate is for the labs. Staying the expert in your own loop is for you. Go be the smartest human in the room, and the machine.
In your corner always,
Patrina, Sarah, Ralph & Jess
Upcoming MSL Mastery Events
 |
 |
|
 |
|
|
 |
|
|
 |
|
|
 |
|
|
For MSL Mastery Members
 |
|
|
 |
|
|
Watch On-Demand
 |
|
|
 |
|
|
Responses