Advertisement

Closed-source software makes for a bad government IT outlook

A former OMB and OSTP policy leader argues that the recent Microsoft outage offered a clear demonstration of the vulnerabilities the government faces due to vendor lock-in.
Listen to this article
0:00
Learn more. This feature uses an automated voice, which may result in occasional errors in pronunciation, tone, or sentiment.
In this photo illustration, an error message is shown when trying to access Microsoft's Outlook on Nov. 12, 2026. Microsoft reported a widespread outage that left users unable to access Outlook, Teams and other Microsoft services. (Photo Illustration by Scott Olson/Getty Images)

When the news broke about the Microsoft Outlook outage concurrently as the Nvidia honeypot scandal was developing, my immediate thought was, “well, Jensen flipped the switch on his way out the door and shut down the internet.” Of course, correlation is not causation but these two emergent crises in the tech sector reveal a critical vulnerability that all of us need eyes on: vendor lock-in.

Vendor lock-in occurs when an organization relies so heavily on a single commercial provider for critical infrastructure that switching to an alternative becomes practically or financially prohibitive. This dependency creates severe vulnerabilities, leaving the organization exposed to the provider’s outages, unpatched security flaws, service throttling, and — often — pricing changes. 

The Microsoft outage puts a sharp focus on the consequences of such vulnerabilities. The outage began around 1 p.m. Eastern time on Thursday, Nov. 12. By that evening, Downdetector had logged nearly 140,000 Outlook problem reports from users in 74 countries, and the disruption ran past the 12-hour mark into Friday. 

Microsoft blamed server problems and said it saw no sign of malicious activity. Governments were among the hardest hit. The U.S. House of Representatives, Japan, Australia’s spy agency, and federal police were all impacted by the disruption. In London, the Metropolitan Police lost the ability to take some crime reports online, while the State of Michigan warned residents that email replies would be delayed. Some organizations even reverted to using fax machines.

Advertisement

In the United States, federal policy already treats this kind of dependence as a risk — and yet agencies are still beholden to the vulnerabilities of vendor lock-in that such outages reveal. 

OMB memorandum M-25-22, an Office of Management and Budget guidance document I helped develop, tells agencies buying AI to watch vendor sourcing, data portability, and long-term interoperability so they don’t end up with costly dependencies on a single vendor. 

It directs them to scope licensing and intellectual property rights to avoid vendor lock-in, and to write lock-in protections — knowledge transfer, data and model portability, rights to code and models produced under the contract — into their AI contracts, building on the information governance requirements of OMB Circular A-130. 

In July, the Cybersecurity and Infrastructure Security Agency made a parallel argument about security. Its guidance on open-source software security points out that open source lets agencies assess code quality directly instead of relying on vendor assurances. It also recommends that when contractors build custom software, the government keeps the rights to reuse, modify, and release that code.

Congress has moved in the same direction, but with a significant gap. The SHARE IT Act, signed into law Dec. 23, 2024, requires agencies to share custom-developed code with one another so taxpayers stop paying twice for the same software. The law covers code written by federal employees and code produced under federal contracts. 

Advertisement

Its definition of custom-developed code excludes commercial computer software and commercial off-the-shelf products, though. Email, calendars, and collaboration suites are typically licensed as commercial cloud services, which puts the systems that went dark Nov. 12 outside the statute entirely. 

That exclusion matters here because Microsoft cannot leverage distributed, independent teams to work on the fix. During the outage, Microsoft said it was trying different methods to find the cause and taking a measured approach to identify and repair affected infrastructure. Meanwhile, all affected agencies can do is wait. Open-source software wouldn’t have prevented a capacity failure — especially a server-side issue like this one — by itself, but it would have let more people hunt for the cause, and it would let agencies lend staff to dig into the problem, too. 

The PHP vulnerabilities disclosed in October show how that distributed model works in practice, flaws and all. PHP runs on about 69% of websites whose server-side language can be identified. Two bugs in the language itself — CVE-2026-109618 and CVE-2026-109619 — allowed unauthenticated remote code execution and were rated 9.2 out of 10 in severity. 

Because the source code is public, several independent teams found them. Ryan Austin, a researcher at the security firm Calif, used an OpenAI model to build a working exploit and reported it on Aug. 11; OpenAI reported the same root cause, and the cloud security company Wiz reported related attack paths later that month. 

The PHP Foundation coordinated a fix with researchers, distributors, and security vendors, then shipped patched versions on Oct. 21, a day ahead of schedule after details leaked. OpenAI rallied defenders through its Patch the Planet program, and CISA urged immediate patching.

Advertisement

The PHP story isn’t a clean win for speed. In its own postmortem, the foundation said a core developer first saw signs of the memory bug in 2024 while fixing an issue from an outside audit funded by Germany’s Sovereign Tech Agency, and that the fix then stalled for months. 

The foundation called that delay a structural failure of its decision process and its resources. We know this because the project published it, along with a dated timeline, credit for each researcher, and the code that fixed the problem. Microsoft’s customers will learn whatever Microsoft decides to put in its post-incident report. The PHP case also shows what public money can buy: transparency and rapid response through investing in the open-source system.

Investors got their own version of this lesson the same week. Markets have concentrated so much capital in one hardware company that Nvidia was worth about $4.5 trillion before the scandal broke, and one executive’s personal life erased roughly $192 billion of that in a single trading session while the rest of the market rose. 

Government IT has made the same bet with its commercial software, routing email, calendars, and meetings for agency after agency through one provider. Both are a lot of eggs in one basket. Open source spreads the risk across many maintainers, many hosts, and many sets of eyes, so one failure doesn’t take everything down with it. 

The lock-in protections in M-25-22 and the code-sharing logic of SHARE IT could be extended to the commercial platforms when they do business with the government, with a preference for open standards and open-source options that would let an agency move its mail when a provider fails.

Advertisement

No single company should be able to take the world offline when it goes down.

Christopher Steven Marcum, PhD is an open science advocate. He previously served in information policy roles at the White House Office of Management and Budget and the White House Office of Science and Technology Policy. 

Christopher Steven Marcum

Written by Christopher Steven Marcum

Christopher Steven Marcum, PhD is an open science advocate. He previously served in information policy roles at the White House Office of Management and Budget and the White House Office of Science and Technology Policy.

Latest Podcasts