The night two critical PHP flaws leaked before their patches, Josh Peterson asked Gary Boyle what was really happening. Gary's answer: AI has made attackers faster than any MSP can patch, the open-weight models they run have no guardrails, and the frontier models defenders use refuse the targeted vulnerability work attackers do freely.

Then Gary turns the question on MSP owners. Most MSPs became brokers of security, handing backup, cloud and antivirus to someone else, and a thousand unpatched endpoints mean something different when the bots never sleep. He argues against freezing AI development and for learning the tools, because AI security expertise is the differentiator MSPs have chased for twenty years.

----------

What We Cover in This Episode

- Unpack the PHP flaws that leaked before their patches - See why security is just code, and why AI can read it - Compare open-weight models with no guardrails to frontier models that refuse - Learn why hackers and developers scanning code do the same thing - Hear what a Bitcoin red team spending $10,000 a day revealed - Rethink patching when a thousand endpoints meet unlimited bots - Decide whether to freeze AI development after a breach - Question whether your MSP is a broker of security - Turn AI security expertise into a real differentiator - Find forgotten API keys before they become a surprise bill

----------

Subscribe & Follow BMK Vision

YouTube (Video Podcast): https://www.youtube.com/@beringmckinleyvision?sub_confirmation=1 Learn More About the Vision Platform: https://beringmckinley.com/vision Apply to Be a Guest: https://beringmckinley.com/blog#speaker-form

----------

SEO Keywords

- AI cyberattacks - MSP cybersecurity - MSP security responsibility - PHP2Shell - AI guardrails - open-weight AI models - API key security - unpatched endpoints - MSP differentiation - managed service providers - BMK Vision Podcast

----------

Credits

Host: Josh Peterson Co-Host: Gary Boyle Producer: Bering McKinley Episode: AI Cyberattacks - Who Owns Your MSP's Security Risk?

Josh Peterson built a working app in Claude in a day without writing a line of code, and he already has clients who could use it. So he asks Gary Boyle the question every MSP owner is starting to ask: why not build tools like this for clients and charge for them?

Gary's answer starts with a map of the AI journey and the stage he calls the danger zone, where a build is exciting, easy to make and hard to support. Together they work out what an MSP has to settle before it sells software nobody on the team fully understands, and the decision every MSP now faces.

----------

What We Cover in This Episode

- Build a working app in Claude without writing code, and see why that changes the MSP's role - Recognize the parallel between selling websites in 1998 and selling AI tools today - Place a business on the AI journey, from patient zero to level three - Spot the danger zone, where a build is easy to make and hard to support - Plan for the day the tech who built it is out and nothing is documented - Set a goal before you build, the same way you would plan a business - See why the first change request is where an unsupported app fails - Decide whether your MSP becomes an AI software company or keeps reselling Copilot

----------

Subscribe & Follow BMK Vision

YouTube (Video Podcast): https://www.youtube.com/@beringmckinleyvision?sub_confirmation=1 Learn More About the Vision Platform: https://beringmckinley.com/vision Apply to Be a Guest: https://beringmckinley.com/blog#speaker-form

----------

SEO Keywords

- MSP AI strategy - sell AI apps to clients - build apps with Claude - AI software development - AI automation for MSPs - AI danger zone - proof of concept - Microsoft Copilot - managed service providers - BMK Vision Podcast

----------

Credits

Host: Josh Peterson Co-Host: Gary Boyle Producer: Bering McKinley Episode: MSP AI Apps - Should You Sell What You Built in Claude?

Josh Peterson gave Claude a ConnectWise API user with write access to his time entries and did not lose a minute of sleep over it. Then he asks Gary Boyle the question every MSP owner is now facing: where is the actual security risk, and how much of it should you accept?

Gary's answer starts with where API keys really leak and ends with a cheat code. An MSP that already trusts ConnectWise and QuickBooks with everything cannot single out the LLM as the bigger risk, you cannot stop data leaving a small business anyway, and the owners who stay AI-first while their competitors lock everything down are the ones pulling ahead. Together they work out what a sane risk assessment looks like, and where an MSP should be on the levels from one to four.

----------

What We Cover in This Episode

- Decide what a Claude integration with ConnectWise actually exposes - Separate one-shotting an app from plan, code and goal - Find out where API keys leak, and why pasting one into an LLM means re-keying - Ask why an MSP that trusts ConnectWise and QuickBooks would fear the LLM more - Recognize when a locked-down Claude has become unusable - Accept that you cannot stop data leaving a small business - Weigh the two forms of risk: leaking data, and holding clients back - Place your MSP on the levels from one to four - Pick two practice projects: dashboards and the PSA-to-accounting connector

----------

Subscribe & Follow BMK Vision

YouTube (Video Podcast): https://www.youtube.com/@beringmckinleyvision?sub_confirmation=1 Learn More About the Vision Platform: https://beringmckinley.com/vision Apply to Be a Guest: https://beringmckinley.com/blog#speaker-form

----------

SEO Keywords

- MSP AI security - AI risk tolerance - Claude for MSPs - ConnectWise API security - API key security - LLM data security - AI governance for MSPs - AI-first MSP - managed service providers - BMK Vision Podcast

----------

Credits

Host: Josh Peterson Co-Host: Gary Boyle Producer: Bering McKinley Episode: MSP AI Security - How Much Risk Should You Accept?