---
title: Who Owns the Risk When an MSP Outsources Security?
description: Who owns the risk when an MSP outsources security? How AI-driven exploits change the answer, and why MSP owners should keep building instead of freezing.
image: https://beringmckinley.com/hubfs/%23131%20-%20AI%20Cyberattacks%20-%20Who%20Owns%20Your%20MSPs%20Security%20Risk_%20_%20BMK%20VISION%20ROUNDTABLE.png
---
[Skip to the main content.](https://beringmckinley.com/blog/who-owns-the-risk-when-an-msp-outsources-security#main-content)
[%20logo.png?width=588&height=164&name=BMK%20(just%20text)%20logo.png "Bering McKinley Logo")](https://beringmckinley.com)
[%20logo.png?width=1212&height=224&name=BMK%20(just%20text)%20logo.png "BMK (just text) logo")](https://beringmckinley.com)
- [What We Do](https://beringmckinley.com/vision)
- [BMK Vision](https://beringmckinley.com/vision)
- [BPA](https://beringmckinley.com/business-performance-assessment)
- [Master Project Plan](https://beringmckinley.com/master-project-plan)
- [Management Consulting](https://beringmckinley.com/management-consulting)
- [Sales Consulting](https://beringmckinley.com/sale-consulting)
- [Application Consulting](https://beringmckinley.com/application-consulting)
- [Finance Consulting](https://beringmckinley.com/finance-consulting)
- [Who We Work With](https://beringmckinley.com/who-we-work-with)
- [MSP / ITSP (1M or Less Revenue)](https://beringmckinley.com/msps-less-than-1m-revenue)
- [MSP / ITSP (1-3M Revenue)](https://beringmckinley.com/msps-1-to-3m-evenue)
- [MSP / ITSP (3-15M+ Revenue)](https://beringmckinley.com/msps-3-to-15m-plus-revenue)
- [Internal IT Departments](https://beringmckinley.com/it-department)
- [Brands We Work With](https://beringmckinley.com/brands-we-work-with)
- [ConnectWise](https://beringmckinley.com/brands-we-work-with)
- [Intuit / QuickBooks](https://beringmckinley.com/brands-we-work-with)
- [MSP Bots](https://beringmckinley.com/brands-we-work-with)
- [IT Glue](https://beringmckinley.com/brands-we-work-with)
- [Datto](https://beringmckinley.com/brands-we-work-with)
- [Lionguard](https://beringmckinley.com/brands-we-work-with)
- [SkyKick](https://beringmckinley.com/brands-we-work-with)
- [Lifecycle Insights](https://beringmckinley.com/brands-we-work-with)
- [Phin](https://beringmckinley.com/brands-we-work-with)
- [Alternative Payments](https://beringmckinley.com/brands-we-work-with)
- [Vendor? Work with us.](https://beringmckinley.com/brands-we-work-with)
- [About](https://beringmckinley.com/about)
- [Team](https://beringmckinley.com/team)
- [Careers](https://beringmckinley.com/careers)
- Other BMK Brands
- [BMK Community](https://bmkcommunity.com)
- [BMK Education](https://bmkcommunity.com/education-and-training/)
- [BMK Operations](https://www.bmkcommunity.com/msp-operations/)
- [Blog](https://beringmckinley.com/blog)
- [Contact](https://beringmckinley.com/contact)
- [Text us](SMS:212.300.3016)
Search
[](https://beringmckinley.com/hs/cta/wi/redirect?encryptedPayload=AVxigLIETtrjUQSedSk%2FevPshVUTCIUTcp6KHoVLuZ0K%2FkWm%2BWrRThQp5wUBTJgYZbrXacgW3l5KTWui0VbOMDbjeme74P9IMPrtaj775nwrPtAMkOQ3xPXjzSN4OqEXbkGXf3FgVpmK2H0wPUJr1mzeE6hVd21EuUt9XmWzAt0Ut7nuvppsuMuHF9uGzc1Zmp4W&webInteractiveContentId=195731585557&portalId=2725629)
Toggle Menu
Search
Toggle Menu
[](https://beringmckinley.com/hs/cta/wi/redirect?encryptedPayload=AVxigLIETtrjUQSedSk%2FevPshVUTCIUTcp6KHoVLuZ0K%2FkWm%2BWrRThQp5wUBTJgYZbrXacgW3l5KTWui0VbOMDbjeme74P9IMPrtaj775nwrPtAMkOQ3xPXjzSN4OqEXbkGXf3FgVpmK2H0wPUJr1mzeE6hVd21EuUt9XmWzAt0Ut7nuvppsuMuHF9uGzc1Zmp4W&webInteractiveContentId=195731585557&portalId=2725629)
- [What We Do](https://beringmckinley.com/vision)
- [BMK Vision](https://beringmckinley.com/vision)
- [BPA](https://beringmckinley.com/business-performance-assessment)
- [Master Project Plan](https://beringmckinley.com/master-project-plan)
- [Management Consulting](https://beringmckinley.com/management-consulting)
- [Sales Consulting](https://beringmckinley.com/sale-consulting)
- [Application Consulting](https://beringmckinley.com/application-consulting)
- [Finance Consulting](https://beringmckinley.com/finance-consulting)
- [Who We Work With](https://beringmckinley.com/who-we-work-with)
- [MSP / ITSP (1M or Less Revenue)](https://beringmckinley.com/msps-less-than-1m-revenue)
- [MSP / ITSP (1-3M Revenue)](https://beringmckinley.com/msps-1-to-3m-evenue)
- [MSP / ITSP (3-15M+ Revenue)](https://beringmckinley.com/msps-3-to-15m-plus-revenue)
- [Internal IT Departments](https://beringmckinley.com/it-department)
- [Brands We Work With](https://beringmckinley.com/brands-we-work-with)
- [ConnectWise](https://beringmckinley.com/brands-we-work-with)
- [Intuit / QuickBooks](https://beringmckinley.com/brands-we-work-with)
- [MSP Bots](https://beringmckinley.com/brands-we-work-with)
- [IT Glue](https://beringmckinley.com/brands-we-work-with)
- [Datto](https://beringmckinley.com/brands-we-work-with)
- [Lionguard](https://beringmckinley.com/brands-we-work-with)
- [SkyKick](https://beringmckinley.com/brands-we-work-with)
- [Lifecycle Insights](https://beringmckinley.com/brands-we-work-with)
- [Phin](https://beringmckinley.com/brands-we-work-with)
- [Alternative Payments](https://beringmckinley.com/brands-we-work-with)
- [Vendor? Work with us.](https://beringmckinley.com/brands-we-work-with)
- [About](https://beringmckinley.com/about)
- [Team](https://beringmckinley.com/team)
- [Careers](https://beringmckinley.com/careers)
- Other BMK Brands
- [BMK Community](https://bmkcommunity.com)
- [BMK Education](https://bmkcommunity.com/education-and-training/)
- [BMK Operations](https://www.bmkcommunity.com/msp-operations/)
- [Blog](https://beringmckinley.com/blog)
- [Contact](https://beringmckinley.com/contact)
- [Text us](SMS:212.300.3016)
- Example Link
- Example Link
- Example Link
[Facebook](https://beringmckinley.com/blog/who-owns-the-risk-when-an-msp-outsources-security#0) [Instagram](https://beringmckinley.com/blog/who-owns-the-risk-when-an-msp-outsources-security#0) [Linkedin](https://beringmckinley.com/blog/who-owns-the-risk-when-an-msp-outsources-security#0) [X](https://beringmckinley.com/blog/who-owns-the-risk-when-an-msp-outsources-security#0) [YouTube](https://beringmckinley.com/blog/who-owns-the-risk-when-an-msp-outsources-security#0) [Medium](https://beringmckinley.com/blog/who-owns-the-risk-when-an-msp-outsources-security#0)
13 min read
# Who Owns the Risk When an MSP Outsources Security?
[ Gary Boyle](https://beringmckinley.com/blog/author/gary-boyle) : Published October 26, 2026
[BMK Vision Roundtable](https://beringmckinley.com/blog/tag/bmk-vision-roundtable)
On Wednesday, October 21, 2026, two critical PHP vulnerabilities leaked online before the patches meant to fix them were released, and the cloud platform Railway told its customers that an unauthorized party may have reached the systems that store their environment variables, which is where most web applications keep their API keys. That evening Josh Peterson asked Gary Boyle for a briefing, and Gary opened with an admission most MSP owners will recognize: he knows enough about security to be super terrified. A few days earlier he had been sitting with a client who was already building AI tools for himself, and when Gary asked what they would build for the client's own customers, the first answer was that he just did not know if it was secure. Gary's reply was a question. We don't really know if anything is secure, do we? The client smiled, because it was true. BMK spent the previous episode on [whether an MSP should trust Claude with client data](https://beringmckinley.com/blog/should-an-msp-trust-claude-with-client-data), which is a question about how much risk to tolerate. This one is about who owns it. It sits on top of two older arguments, about why [data risk is the MSP's invisible liability](https://beringmckinley.com/blog/data-risk-is-the-msps-invisible-liability) and about the [cybersecurity practices that protect MSP clients](https://beringmckinley.com/blog/key-best-practices-for-msp-cybersecurity-protecting-clients), and it asks the question both of them leave open: who owns the risk when an MSP outsources security, and what changes when the attackers are running AI?
Gary's answer is not comfortable, and BMK thinks it is right. The leadership judgment is that an MSP has to decide how much of the risk it actually understands, because outsourcing the tooling has never outsourced the client's expectation that you own the outcome. The execution discipline is old and newly urgent: patch what is exposed, rotate the keys you forgot you created, and stop treating "not a problem yet" as a security policy now that software can probe every server on the internet without getting tired. The positioning question is what an MSP becomes when its entire security story is a list of vendors, and Gary has a phrase for it: a broker of security. The financial leverage runs in both directions. A leaked API key can turn into a surprise bill for a few thousand dollars, and a client's neglected 2015 website is suddenly a funded project with a security reason attached. And the long-term strategy is the part that should change how an owner spends the next year. The same AI tools the attackers use are available to every MSP, and the owner who puts in the time to understand them holds the differentiator this industry has been chasing for twenty years. What follows is how BMK thinks about each of those.
---
## Listen on Your Favorite Platform
- [](https://podcasts.apple.com/us/podcast/the-bmk-vision-podcast/id1827082235)
- [](https://www.youtube.com/@beringmckinleyvision?sub_confirmation=1)
- [](https://music.youtube.com/playlist?list=PL03jvZSuXCDFHzZFwDdTs1xP07b8AI_5P)
- [](https://open.spotify.com/show/6iYx2sDBy2YnN3V2YA9GyW)
- [](https://music.amazon.com/podcasts/33fb1a31-0a2b-4478-80b9-38fc3c552675/the-bmk-vision-podcast)
- [](https://www.podbean.com/pw/pbblog-k23bf-14140db?from=usersite)
- [](https://www.iheart.com/podcast/269-the-bmk-vision-podcast-259005021/)
- [](https://player.fm/series/the-bmk-vision-podcast)
- [](https://www.listennotes.com/podcasts/the-bmk-vision-podcast-josh-peterson-7iS7AIHHUhL/)
- [](https://www.podchaser.com/podcasts/the-bmk-vision-podcast-5871877)
- [](https://www.boomplaymusic.com/podcasts/136263)
---
## The Night the Exploits Got Faster Than the Patches
Gary's briefing started with a number. By his estimate, about 70 percent of the web runs on PHP, the language under most of the WordPress sites built through the 2010s. The two flaws that leaked on October 21, tracked as CVE-2026-109618 and CVE-2026-109619 and nicknamed PHP2Shell, were rated 9.2 out of 10 for severity, and public reporting said they could let an attacker run code on an affected server without logging in, on PHP versions going back to PHP 5. The PHP project had planned to ship fixes at 14:00 UTC that day. The details got out first, and security firms reported exploitation attempts before the patches landed. The same day, Railway, a cloud platform that hosts web applications, disclosed that an unauthorized party may have accessed customers' encrypted environment variables and, for some customers, their decrypted contents. It asked every user to rotate their secrets, API keys included, and said it believed the attack was part of a series aimed at platform companies in order to reach AI inference providers. Gary's read was that none of this is a single event. It is what the past year has looked like since language models got good enough to read code deeply, because, as he put it, security is just code. An exploit is a flaw in code, and a model that can write code can assess it.
His most useful observation for an MSP owner was about patching. Coders, he said, are fairly lazy, and MSPs know it as well as anyone. You have a thousand endpoints out there, and most of them are not patched on the day the patch comes out. There are good reasons for that, and underneath all of them is the same one: you do not have time to fix a non-problem. That policy was rational in a world where nobody was actively trying to exploit every unpatched system at once. With AI, Gary argued, the number of bots an attacker can fire at vulnerable systems is effectively unlimited, and an exploit prompt is no harder to write than a prompt for tonight's dinner recipe. The model is the intelligence layer most attackers used to be missing. That is what he meant by a step-function change. The vulnerabilities are not new. The cost of finding and using them has collapsed.
- A leaked flaw can be exploited before its patch ships. Your patch window is now measured against software that never sleeps.
- "It isn't a problem yet" was a defensible patching policy only while nobody was probing every server at once.
- API keys stored as environment variables are a target in their own right. Know where yours live before someone else does.
## Attackers Get Unguarded Models. Defenders Get Refusals.
Halfway through the conversation, Josh ran a test of his own. He pasted a block of code into Claude, asked whether anything in it was exploitable, and got back an answer that the code was rock solid, followed by a paragraph he did not understand. Then he asked why it had agreed to look at all, since Gary had just said these models will not examine code, and Claude pushed back on the premise. Gary's response was the most important distinction of the episode. A model will review code in a general way. What the US frontier labs gate is the targeted request: assess this code for these specific vulnerabilities. The reason is that the job is identical on both sides. A hacker scans code to find out whether it is vulnerable so he can exploit it. A developer scans code to find out whether it is vulnerable so he can protect it. Same function. The labs do not want the liability of handing that function to whoever asks, so they wrap it in guardrails, and a vague question like "is my code secure" gets a confident answer that is not an assessment. Gary's blunt version was that the model is essentially lying to you, because you never gave it the context for what it was supposed to check.
The people who learned this first were in Bitcoin. After a wallet vulnerability let attackers drain funds this summer, a volunteer effort that includes Rob Hamilton, the CEO of AnchorWatch, began scanning open-source Bitcoin repositories with AI. Published accounts describe the group, known as Bitcoin Red Team, using models such as Moonshot's Kimi K3 and Z.ai's GLM 5.2 alongside American models, spending on the order of $10,000 a day, and running into restrictions on security analysis even after completing a lab's identity verification. Gary drew the lesson an MSP owner should draw from it. The guardrails exist for the general public, and credentialed access to unguarded models is slow to get. Meanwhile, anyone can load an open-weight model with no guardrails into an open-source agent harness and do the work the frontier models decline. The red team proved the capability was real, which helped them work out what was happening, and in the same stroke it showed every attacker a playbook. That asymmetry is the point. The people probing your clients are not waiting on anyone's credentialing program.
- "Is my code secure?" is not a security test. Name what you are checking for, or the answer is decoration.
- The capability that finds a flaw is the capability that exploits it, which is why the labs gate it and why attackers route around them.
- Plan as if the other side has unrestricted tools, because it does.
## Who Owns the Risk When an MSP Outsources Security?
Gary's sharpest line came right after Josh described a copier dealer he once worked with: fifty to seventy-five million dollars in copier revenue, two or three million in a young MSP, and a board whose average age was well over sixty-five telling the founder to get out. Why take the risk? We're all going to die rich. Gary's answer started with how much an owner internalizes that risk: how much he cares about his clients being up and secure, feeding their families, keeping their employees employed. Then he described what MSPs actually did as the job got harder. They outsourced it. The people outsourcing to the MSP had hired it precisely so that everything would be taken care of, and the MSP turned around and said someone else needs to take that on. Someone else handles my backup, my cloud, my antivirus management. Someone else figures out how to make my systems secure so I can sell more of them. That is what an MSP is, Gary said: a broker of security. Very few do high-level security work themselves. They do their best, and until they get hit, they feel okay.
Josh supplied the definition that turns this from a confession into a strategic question. Years ago, at a conference, someone defined managed services as the transfer of risk and responsibility for a network from its owner to a third party. By that definition, an MSP that passes the risk on again, to a fourth, fifth and sixth party, has done nothing wrong. It has done the job as designed. BMK's view is that the definition is incomplete in a way that matters now. You can transfer the work. On paper, you can transfer a good deal of the liability. What you cannot transfer is the client's belief about who is responsible, and when a client's servers are encrypted, that belief arrives at your door, not at your vendor's. Gary put it to every owner listening as a question: if your client got hacked and got ransomware on their servers, would you feel responsible? Would you go home and really lose sleep over that? He did, and it is a large part of why he sold his own MSP. The honest answer an owner gives to that question is the real answer to who owns the risk.
- You can outsource the tooling. You cannot outsource the client's expectation that you own the outcome.
- Every additional layer of outsourcing adds a vendor whose response time you do not control, on the day you most need it.
- Decide deliberately how much of the risk you will understand yourself. That is the one part no vendor can sell you.
## Should an MSP Freeze AI Development After a Breach?
Josh opened the episode with two clients who had both been working at the edge of AI. The first, led by an owner under forty, has been cautious about security and slow to move. The second is led by an owner in his sixties, if not seventy, who connected Claude to ConnectWise and within forty-eight hours had built himself dashboards, pulled new insights out of his own data, and made a valuation calculator. Then Josh named the reaction he could see coming after a night like October 21: freeze any development that faces the outside world, because either we know what we know and freezing feels like the best move, or we don't know what we don't know. Gary's answer was no. He separated two questions that tend to get welded together. Should an MSP build software with AI, for itself and for its clients? Yes, and it is still a massive opportunity. Can anyone keep the internet secure in a world of AI-driven attacks, whether that is Microsoft, the antivirus vendors or the cloud platforms? That is totally unclear, and the question exists whether or not you write a line of code. Someone is going to develop the software. Your clients already carry the exposure.
The opportunity Gary pointed to is unglamorous and real. Every client with an informational website built around 2015, often for something like thirty thousand dollars and running on an old stack nobody has patched, now has a reason to replace it. A rebuild on newer, more resilient infrastructure might cost three thousand dollars and be ten times better, and the security argument makes the case for the client. Josh's copier dealer board was making a legitimate choice when it asked why take the risk, and Gary's own decision to sell his MSP was the same choice, made honestly. BMK's view is that the choice is valid in either direction. What is not valid is making it by default, freezing because the news was frightening rather than because you decided you do not want the responsibility. Josh's version of the conclusion was simple: if somebody is going to build it, the MSP that brings a security lens to the work is the right one to do it.
- Freezing your own development does not reduce your clients' exposure. Their old sites and servers are already the attack surface.
- The neglected website rebuild is a real project with a security reason attached, and it can be sold now.
- Choosing not to carry the risk is legitimate. Drifting into that choice because the headlines were bad is not.
## The Differentiator MSPs Have Chased for Twenty Years
Gary's model for what an owner should do next was Rob Hamilton. In fifteen to twenty years around Bitcoin development, Hamilton had never submitted code to Bitcoin Core. He could build software, run a business and understand advanced cryptography, and by his own account he did not really know security until he started using the newer models. Software development and cybersecurity are not the same discipline, Gary said, any more than running an MSP is. Most developers know nothing about security, just like most MSPs. They know best practices, they run the vulnerability scans and the pen tests, and they accept what the tools tell them. Working unpaid, and by Gary's account paying for the tokens himself, Hamilton scanned repositories with AI for a month or two and came out with real expertise in one specific kind of security. There was no CEO coming to fix the problem, Gary said, so he rolled up his sleeves. Josh compared it to a flood: the National Guard and the local fire department can only do so much, and the people who matter are the ones out bagging sand. Gary's point was that this is all an MSP is, too. It comes down to how much you want to learn and how much time you are willing to put in the seat.
Then he made the business case. MSPs have spent two decades trying to answer why they are different, and the honest answer has usually been that they are not. Everyone sells the same products under the same agreements at about the same price, and the differentiator on the website says good service. An MSP that can credibly say it understands AI security, and can build secure, custom applications that do what a client actually needs instead of bending an off-the-shelf tool to fit, has something its competitors cannot buy from a distributor. It starts at home. Gary closed with a warning about the most ordinary consequence of a week like this one: plenty of people have API keys sitting in places they have forgotten about, and someone is going to get a bill for a few thousand dollars and wonder why their little app cost so much. The MSP that knows where every one of its own keys lives, and can teach its clients to do the same, has already started building the expertise Gary is describing.
- Security judgment comes from time in the seat. AI compresses that time for anyone willing to spend it.
- AI security expertise plus secure custom applications is a position a competitor cannot copy by buying the same tool.
- Start with your own exposure: inventory every API key, where it lives, and who could run up a bill with it.
## Frequently Asked Questions
### Who owns the risk when an MSP outsources security?
An MSP can transfer the work, and some of the contractual liability, to its vendors. It cannot transfer the client's expectation that the MSP owns the outcome, and when a client is breached, the call comes to the MSP. The practical answer is to decide deliberately how much of the risk the MSP will understand itself, choose vendors whose response it can verify, and be honest with clients about where responsibility actually sits.
### Should MSPs stop building software with AI after a major breach?
Not by default. Freezing an MSP's own development does not reduce the exposure its clients already carry on old websites and unpatched servers. The better response is to keep building with a security lens: plan the application, know where its credentials live, and treat security as part of the work rather than a reason to stop.
### Why won't AI tools run a full security audit of my code?
Frontier models will review code, but requests framed as finding exploitable vulnerabilities can run into safety restrictions, because the same analysis serves attackers and defenders. A vague prompt such as "is my code secure" tends to produce a confident answer that is not a real assessment. Be specific about what you are testing for, and pair AI review with purpose-built security tools and human expertise.
### What were the PHP2Shell vulnerabilities?
PHP2Shell is the name given to two critical PHP vulnerabilities, CVE-2026-109618 and CVE-2026-109619, rated 9.2 for severity, which public reporting said could allow remote code execution without authentication on PHP 5 and later. Details leaked ahead of the coordinated patch release on October 21, 2026, and exploitation attempts were reported before the fixes landed. Any server still running an unpatched PHP version should be treated as exposed.
### What should an MSP do if its API keys may have been exposed?
Rotate every key and credential that may have been exposed, revoke the old ones with the provider, and review logs and usage for activity you do not recognize. Then build a list of where every key lives. Forgotten keys are the ones that turn into surprise usage bills, sometimes for thousands of dollars.
### How can an MSP differentiate itself in the AI era?
By building expertise its competitors cannot buy. Most MSPs sell the same products under similar agreements at similar prices. An MSP that understands AI security and can deliver secure, custom applications for its clients offers something that cannot be copied by switching vendors.
## Episode Highlights
- **00:00** - Cold open: MSPs became brokers of security, so would you lose sleep if a client got ransomware?
- **00:41** - Two MSPs, two speeds: a cautious owner under forty, and an owner near seventy building with Claude in 48 hours.
- **02:15** - Two critical PHP flaws leaked before their patches, on a language much of the web still runs.
- **04:42** - Most MSP owners know enough about security to be terrified, and nobody really knows if anything is secure.
- **06:49** - Denial of service, data theft or ransomware? All of the above, and the response is already global.
- **08:17** - Security is just code, and the open-weight models attackers use come with no guardrails.
- **10:20** - A general code review works, but targeted vulnerability scans hit guardrails because hackers and developers do the same thing.
- **11:02** - A Bitcoin wallet drain led a volunteer red team to spend about $10,000 a day scanning code with AI.
- **15:10** - Page source is harmless; the exploits live in the language, the server and the database behind it.
- **16:23** - Railway's exposed environment variables show why API keys stored in the cloud are a target.
- **17:46** - An exploit prompt is no harder than a dinner recipe prompt, because AI is the intelligence layer hackers lacked.
- **18:55** - Josh's live test: Claude calls his code rock solid, and Gary explains why a vague question gets a confident answer.
- **21:47** - What consumers will see next: unpatched websites going offline, in an attack that feels more decentralized than past ransomware waves.
- **23:16** - A step-function change: a thousand unpatched endpoints meet an unlimited supply of AI-driven bots.
- **25:55** - Should MSPs freeze any AI development that faces the outside world?
- **26:39** - Gary says keep building, starting with the $30,000 website from 2015 that nobody has patched.
- **29:29** - A copier dealer's board asks why take the risk, and whether an owner has the stomach for the responsibility.
- **31:17** - MSPs became brokers of security by handing every hard problem to someone else.
- **33:10** - Ransomware funds terrorism, and managed services as the transfer of risk and responsibility to a third party.
- **34:46** - Moving back to the server room will not help; the same AI tool is in everyone's hands, so learn it.
- **38:42** - A 2016 warning from a security-minded MSP reads differently ten years later.
- **40:29** - Rob Hamilton's story: development skill is not security skill, and AI can close that gap in months.
- **45:35** - The differentiator MSPs have chased for twenty years: AI security expertise and secure custom applications.
- **47:35** - Forgotten API keys will produce surprise bills of a few thousand dollars.
## About the Co-Host: Gary Boyle
Gary Boyle is a Partner for Strategy & Business Development at Bering McKinley. With a background spanning network engineering, entrepreneurship, and strategic consulting, Gary brings real-world operator experience to helping MSP owners build stronger, more profitable businesses. On the Vision roundtable he co-hosts with Josh Peterson, working through the decisions that determine how an MSP grows, sells and runs.
[Connect with Gary on LinkedIn →](https://www.linkedin.com/in/garyboyle/)
## About the Host: Josh Peterson
Josh Peterson is the CEO of Bering McKinley and host of The BMK Vision Podcast. Since 2004, Josh has worked with hundreds of MSP owners to build operationally sound, profitable businesses through consulting, peer teams, and direct coaching.
[Connect with Josh Peterson on LinkedIn →](https://www.linkedin.com/in/joshdpeterson/)
## Related Resources from Bering McKinley
- [MSPs are being commoditized: here's the way out](https://beringmckinley.com/blog/msps-are-being-commoditized-heres-the-way-out)
- [Should an MSP build its own software instead of paying for it?](https://beringmckinley.com/blog/should-an-msp-build-its-own-software-instead-of-paying-for-it)
## Want to Continue the Conversation?
Deciding how much security risk your MSP owns, and how much it passes on, is a strategic decision before it is a vendor decision. Making that call deliberately, with goals and a plan behind it, is what the Vision Operating System is built to do.
[Apply to be a Guest](https://beringmckinley.com/blog#speaker-form) [Learn More About Vision](https://beringmckinley.com/vision)
- [Tweet](https://twitter.com/share)
Menu
Menu
- [What We Do](https://beringmckinley.com/vision)
- [Who We Work With](https://beringmckinley.com/who-we-work-with)
- [Brands We Work With](https://beringmckinley.com/brands-we-work-with)
- [About](https://beringmckinley.com/about)
- Blog
- [Contact](https://beringmckinley.com/contact)
More from BMK
More from BMK
[BMK Education](https://education.beringmckinley.com/)
[BMK Community](https://bmkcommunity.com/)
[BMK Ops](https://bmkoperations.com/)
Bering McKinley
Bering McKinley
650 Warrenville Road
Suite 100 #1007
Lisle, IL 60532
Phone: [212.300.3016](tel:212.300.3016)
[Text us: 212.300.3016](sms:212.300.3016)
Email: [info@beringmckinley.com](mailto:info@beringmckinley.com)
[Click here to schedule a meeting](https://app.hubspot.com/meetings/joshpeterson/introduction-to-bering-mckinley)
[%20logo.png)](https://beringmckinley.com)
- [Policies, Terms and Conditions](https://beringmckinley.com/policy)
© 2026 All rights reserved. Bering McKinley
[Facebook](https://www.facebook.com/BeringMckinley)[X](https://twitter.com/beringmckinley)[Instagram](https://www.instagram.com/beringmckinley/?hl=en)[Linkedin](https://www.linkedin.com/company/bering-mckinley)
```json
{
"@context" : "https://schema.org",
"@type" : "BlogPosting",
"author" : {
"@type" : "Person",
"name" : "Gary Boyle",
"url" : "https://beringmckinley.com/blog/author/gary-boyle"
},
"dateModified" : "2026-10-26T12:00:01.359Z",
"datePublished" : "2026-10-26T12:00:01.000Z",
"headline" : "Who Owns the Risk When an MSP Outsources Security?",
"image" : [ "https://beringmckinley.com/hubfs/%23131%20-%20AI%20Cyberattacks%20-%20Who%20Owns%20Your%20MSPs%20Security%20Risk_%20_%20BMK%20VISION%20ROUNDTABLE.png" ],
"mainEntityOfPage" : {
"@id" : "https://beringmckinley.com/blog/who-owns-the-risk-when-an-msp-outsources-security",
"@type" : "WebPage"
},
"publisher" : {
"@type" : "Organization",
"logo" : {
"@type" : "ImageObject",
"url" : "https://beringmckinley.com/hubfs/BMK%20(just%20text)%20logo.png"
},
"name" : "Bering McKinley"
}
}
```
```json
{
"@context" : "https://schema.org",
"@type" : "Article",
"about" : [ {
"@type" : "Thing",
"name" : "MSP security responsibility"
}, {
"@type" : "Thing",
"name" : "AI-driven cyberattacks"
}, {
"@type" : "Thing",
"name" : "AI guardrails and vulnerability scanning"
}, {
"@type" : "Thing",
"name" : "API key security"
} ],
"articleSection" : [ "Listen on Your Favorite Platform", "The Night the Exploits Got Faster Than the Patches", "Attackers Get Unguarded Models. Defenders Get Refusals.", "Who Owns the Risk When an MSP Outsources Security?", "Should an MSP Freeze AI Development After a Breach?", "The Differentiator MSPs Have Chased for Twenty Years", "Frequently Asked Questions", "Episode Highlights", "About the Co-Host: Gary Boyle", "About the Host: Josh Peterson", "Related Resources from Bering McKinley", "Want to Continue the Conversation?" ],
"author" : [ {
"@type" : "Person",
"jobTitle" : "CEO",
"name" : "Josh Peterson",
"url" : "https://www.linkedin.com/in/joshdpeterson/",
"worksFor" : {
"@type" : "Organization",
"name" : "Bering McKinley",
"url" : "https://beringmckinley.com"
}
}, {
"@type" : "Person",
"jobTitle" : "Partner, Strategy & Business Development",
"name" : "Gary Boyle",
"url" : "https://www.linkedin.com/in/garyboyle/",
"worksFor" : {
"@type" : "Organization",
"name" : "Bering McKinley",
"url" : "https://beringmckinley.com"
}
} ],
"dateCreated" : "2026-10-23T08:00:00-04:00",
"dateModified" : "2026-10-26T08:00:00-04:00",
"datePublished" : "2026-10-26T08:00:00-04:00",
"description" : "Who owns the risk when an MSP outsources security? How AI-driven exploits change the answer, and why MSP owners should keep building instead of freezing.",
"headline" : "Who Owns the Risk When an MSP Outsources Security?",
"image" : {
"@type" : "ImageObject",
"url" : "https://2725629.fs1.hubspotusercontent-na1.net/hub/2725629/hubfs/%23131%20-%20AI%20Cyberattacks%20-%20Who%20Owns%20Your%20MSPs%20Security%20Risk_%20_%20BMK%20VISION%20ROUNDTABLE.png?width=2000&name=%23131%20-%20AI%20Cyberattacks%20-%20Who%20Owns%20Your%20MSPs%20Security%20Risk_%20_%20BMK%20VISION%20ROUNDTABLE.png"
},
"inLanguage" : "en",
"isPartOf" : {
"@type" : "WebSite",
"name" : "Bering McKinley Blog",
"url" : "https://beringmckinley.com/blog/"
},
"keywords" : [ "who owns the risk when an MSP outsources security", "MSP security responsibility", "AI cyberattacks", "PHP2Shell", "AI guardrails", "open-weight AI models", "API key security", "unpatched endpoints", "MSP differentiation", "AI security for MSPs", "Vision Operating System", "managed service provider" ],
"mainEntityOfPage" : {
"@id" : "https://beringmckinley.com/blog/who-owns-the-risk-when-an-msp-outsources-security",
"@type" : "WebPage"
},
"mentions" : [ {
"@type" : "Thing",
"name" : "Vision Operating System"
}, {
"@type" : "Person",
"name" : "Gary Boyle"
}, {
"@type" : "Person",
"name" : "Josh Peterson"
}, {
"@type" : "Organization",
"name" : "Bering McKinley"
}, {
"@type" : "SoftwareApplication",
"name" : "Claude"
}, {
"@type" : "ComputerLanguage",
"name" : "PHP"
}, {
"@type" : "Person",
"name" : "Rob Hamilton"
} ],
"name" : "Who Owns the Risk When an MSP Outsources Security?",
"potentialAction" : [ {
"@type" : "ReadAction",
"target" : "https://beringmckinley.com/blog/who-owns-the-risk-when-an-msp-outsources-security"
}, {
"@type" : "WatchAction",
"target" : "https://youtu.be/QCwVhvkMPWc"
} ],
"publisher" : {
"@type" : "Organization",
"logo" : {
"@type" : "ImageObject",
"url" : "https://beringmckinley.com/hs-fs/hubfs/BMK%20(just%20text)%20logo.png"
},
"name" : "Bering McKinley",
"url" : "https://beringmckinley.com"
},
"speakable" : {
"@type" : "SpeakableSpecification",
"cssSelector" : [ "div > div > p:first-of-type", "div > div > p:nth-of-type(2)" ]
},
"timeRequired" : "PT17M",
"url" : "https://beringmckinley.com/blog/who-owns-the-risk-when-an-msp-outsources-security",
"video" : {
"@type" : "VideoObject",
"contentUrl" : "https://youtu.be/QCwVhvkMPWc",
"description" : "Josh Peterson and Gary Boyle on AI cyberattacks and who owns an MSP's security risk, from exploits that outrun patching and guardrails that bind defenders but not attackers to why AI security expertise is the differentiator MSPs need.",
"duration" : "PT48M46S",
"embedUrl" : "https://www.youtube.com/embed/QCwVhvkMPWc",
"name" : "#131 - AI Cyberattacks - Who Owns Your MSP's Security Risk? | BMK VISION ROUNDTABLE",
"publisher" : {
"@type" : "Organization",
"name" : "Bering McKinley",
"url" : "https://beringmckinley.com"
},
"thumbnailUrl" : "https://img.youtube.com/vi/QCwVhvkMPWc/maxresdefault.jpg",
"uploadDate" : "2026-10-26T08:00:00-04:00"
},
"wordCount" : "3801"
}
```
```json
{
"@context" : "https://schema.org",
"@type" : "BlogPosting",
"author" : [ {
"@type" : "Person",
"name" : "Gary Boyle",
"url" : "https://beringmckinley.com/blog/author/gary-boyle"
} ],
"dateModified" : "2026-10-26T12:00:01Z",
"datePublished" : "2026-10-26T12:00:01Z",
"description" : "Who owns the risk when an MSP outsources security? How AI-driven exploits change the answer, and why MSP owners should keep building instead of freezing.",
"headline" : "Who Owns the Risk When an MSP Outsources Security?",
"image" : [ "https://2725629.fs1.hubspotusercontent-na1.net/hubfs/2725629/%23131%20-%20AI%20Cyberattacks%20-%20Who%20Owns%20Your%20MSPs%20Security%20Risk_%20_%20BMK%20VISION%20ROUNDTABLE.png" ],
"inLanguage" : "en",
"keywords" : "BMK Vision Roundtable",
"mainEntityOfPage" : {
"@id" : "https://beringmckinley.com/blog/who-owns-the-risk-when-an-msp-outsources-security",
"@type" : "WebPage"
},
"publisher" : {
"@type" : "Organization",
"name" : "Bering McKinley"
}
}
```