One Tag Turns Into Seven
Your security team reviewed the marketing tag. They approved it. They moved on.
That tag loaded another script. That script loaded three more. Within a few hops, your checkout page is executing code from vendors nobody on your team has ever heard of. Those scripts have the same access to your forms, your payment fields, and your customer data as the code your own engineers wrote.
This is the approval gap. The distance between what security signed off on and what is actually running on your site right now.
[INSIGHT] 64% of third-party applications access sensitive data without business justification, up from 51% in 2024. The gap is widening, not closing.
It is not a hypothetical anymore. Research from Reflectiz analyzing 4,700 leading websites found that nearly two-thirds of third-party apps on the web touch sensitive data they have no reason to access. Marketing tools are the primary driver.
The problem is structural. Marketing teams deploy tags quickly because they value speed. Security teams review code carefully because they value thoroughness. The scripts that appear in the gap between those two priorities belong to no one. That is exactly why they slip past firewalls, WAFs, and point-in-time code review.
Marketing Owns Most of the Risk
The data tells an uncomfortable story about who is creating the exposure.
Marketing and digital departments now drive 43% of all third-party risk exposure on the web. IT accounts for just 19%. The people least likely to understand script security are the ones adding the most scripts.
This is not a knock on marketers. They are doing their jobs. Conversion tracking, retargeting pixels, chat widgets, personalization tools, A/B testing platforms. Each one needs a tag. Each one needs access to data. Each one was probably installed by someone who never thought about what happens when that tag calls home and brings friends.
[INSIGHT] Marketing and digital teams drive 43% of third-party risk exposure. IT drives 19%. The people adding the most scripts have the least security oversight.
Forty-seven percent of apps running in payment frames lack business justification. Marketing teams frequently deploy conversion tools into these sensitive environments without realizing what they are touching. A pixel on a checkout page is not the same as a pixel on a blog post. One tracks a reader. The other sits inside your payment flow with access to card numbers, billing addresses, and customer credentials.
We have seen this governance gap before in the AI agent space. The same pattern is playing out with web tags, except the blast radius is your entire customer data surface.
AI Made the Gap Wider
Here is where the problem gets worse in 2026.
AI-driven ad tech spins up new integrations, endpoints, and data flows at machine speed. The approval you granted last quarter describes a stack that no longer exists. Your vendor added a new partner integration last Tuesday. That partner added a data-sharing agreement on Thursday. By Monday, your approved tag is loading code from a company that did not exist when you ran your security review.
as The Hacker News reported in their coverage of the approval gap, a script that was clean at approval can change the day after. AI is also making browser abuse cheaper, faster, and accessible even to non-technical attackers.
The scale compounds. Taboola alone reaches 600 million daily active users across 9,000 publisher partners. When an ad tech platform of that size updates its code, the ripple hits thousands of sites simultaneously. Multiply that across every ad tech vendor in your tag manager, and you have a constantly shifting attack surface that no point-in-time audit can capture.
The Awareness Action Gap
Security leaders know this is a problem. They just are not doing anything about it.
Eighty-one percent of security leaders call web attacks a top priority. Only 39% have deployed solutions to address them. That is a 42-point gap between knowing and acting.
The obstacles are familiar. Budget constraints at 34%. Regulation complexity at 32%. Staffing shortages at 31%. The combination hits hardest in the public sector, where government sites saw malicious activity spike from 2% to 12.9% and education sites now show active compromise on 1 in 7 domains.
Insurance, by contrast, reduced malicious activity by 60% by investing in governance and monitoring. The difference is not technical sophistication. It is budget and priority.
[INSIGHT] 81% of security leaders prioritize web attacks. Only 39% have deployed solutions. The 42-point gap explains why unjustified access grows 25% year over year.
The sectors that invest in continuous monitoring are stabilizing. The ones that treat security as a one-time review are losing ground. This mirrors what we documented about the authorization liability gap in AI agent deployments. The pattern is the same: approval without continuous monitoring is not security. It is paperwork.
The Compliance Clock Is Already Running
If the security argument does not move you, the compliance one should.
PCI DSS 4.0.1 Requirements 6.4.3 and 11.6.1 became mandatory on March 31, 2025. They require organizations to inventory all scripts executing on payment pages, ensure they are authorized and justified, and monitor for unauthorized changes. If your marketing tag is loading fourth-party code on your checkout page and you cannot account for it, you are already non-compliant.
The implementation guide from Feroot walks through the specifics, but the headline is simple: you must know what is running on your payment pages at all times. Not what you approved. What is actually executing in your users' browsers right now.
GDPR and CCPA add another layer. Every unvetted script that touches user data is a potential violation. The 64% of apps with unjustified access are not just security risks. They are regulatory liabilities waiting to be discovered during an audit or, worse, after a breach.
This is the same vendor liability trap we covered in the AI space, but the stakes are more immediate. Web tags are live right now. They are not a future risk. They are a present exposure.
The Facebook Pixel Problem
If you want a single example that captures the scale of this risk, look at the Facebook Pixel.
It runs on 53.2% of the web. That is not a typo. More than half of all major websites have the Facebook Pixel installed. In many deployments, it has been configured with Full DOM Access and Automatic Advanced Matching, which transforms it from a marketing tool into a data scraper that captures everything a user types on a page.
If the Facebook Pixel were compromised, the scale would be roughly five times larger than the 2024 Polyfill.io attack. Polyfill affected 100,000 sites over weeks. A Facebook Pixel compromise would hit 2.5 million sites instantly. The data exposed would include form inputs, payment fields, and personal identifiers across half the commercial web.
The fix is not to remove the pixel. It is to scope it. Restrict pixels to landing pages where they serve their actual purpose. Block them from payment frames, credential pages, and any form that collects sensitive data. A pixel that cannot justify its presence on a page should not be on that page.
What Actually Works
The organizations that perform well in the Reflectiz research share a few traits. They maintain fewer than eight third-party apps while average sites struggle with 15 to 25. They monitor continuously rather than auditing once. And they have a joint review process where marketing and security both own the tag deployment pipeline.
Three steps to start closing the gap.
First, inventory every pixel and tracker on your site. Identify who owns it and what business justification it has for being there. Remove anything that cannot answer those two questions. Priority fixes: disable Facebook Pixel's Automatic Advanced Matching on PII pages, verify Google Tag Manager has no payment page access, and review every Shopify app's permissions.
Second, deploy runtime monitoring. You need automated detection for sensitive field access, real-time alerts for unauthorized script behavior, and CSP violation tracking. Point-in-time review is not enough when scripts can change their behavior remotely.
Third, fix the marketing-IT divide. Joint reviews between CISO and CMO leadership for any tool touching payment frames. Tag manager scoping with allow and exclusion lists. And ROI versus security risk assessments before any new tag goes live.
The Part Nobody Wants to Hear
The hardest part of this problem is not technical. It is organizational.
Marketing owns the tag manager in most companies. Security owns the risk assessment. Nobody owns the gap between them. The scripts that load in that gap belong to no one, which is exactly why they are never reviewed, never scoped, and never removed until something breaks.
The companies that fix this do not do it with better tools. They do it with a clear ownership model where every script on every page has a named owner, a documented justification, and a continuous monitoring plan. The technology is solvable. The politics are harder.
The bad data that trains AI bidding to waste your ad budget starts with the same root cause: nobody owns the pipeline between what gets deployed and what gets monitored. Tags are the same story on a different surface.
Your marketing tags are already live on your pages. Do you know what they are actually doing right now?