As an adult, a father, and a data scientist, accountability plays a role in almost every part of my life, from the decisions I make, the things I say, and the work I produce.
We’re all accountable to someone or something, whether that's our families, our coworkers, our communities, our own conscience. And that accountability exists for a reason. It is one of the foundations of trust. If I build a machine learning model at work that creates a bad outcome for an executive, board member, or important customer, I’m going to hear about it. If I let a curse word slip around my toddler or drive a little too fast, there may be consequences there too.
But as AI systems become more capable and more influential, that same sense of accountability can become surprisingly hard to pin down. The problem isn’t that accountability is impossible. It’s that making it technical, portable, and verifiable is still harder than it should be.
That’s part of what we’re trying to solve with BlkSeal.
AI Is Gaining Authority Faster Than Accountability
AI is moving quickly from something we experiment with to something businesses and consumers actually rely on. It writes content, engages customers, produces reports, recommends decisions, analyzes data, interacts with other systems, and increasingly acts through autonomous agents.
The amount of information being created is growing right along with it. IDC estimates that enterprises created 6.9 petabytes of data every second in 2025, and expects that figure to reach 17.1 petabytes per second by 2029, driven in part by agentic AI. [1]
AI is also becoming a larger part of the information we see and consume. Pew Research Center recently analyzed 490,000 English-language webpages and found that, among pages with publication dates after ChatGPT's release, 35% showed significant signs of AI authorship or editing in its July 2026 sample. [2]
And businesses aren't sitting on the sidelines. McKinsey's 2025 State of AI research found that 88% of respondents said their organizations were using AI in at least one business function, up from 78% the year before. [3]
So the direction is pretty clear. We're creating more data. More of it is being generated or influenced by AI. And businesses are putting AI into more of the systems and workflows that people depend on.
Unfortunately, trust isn't scaling at the same rate.
The problem is that adoption and trust are moving at very different speeds. Edelman's 2025 AI Trust Barometer found that only 32% of Americans said they trust AI. [4] Pew found something similar from another angle. In early 2026, 59% of U.S. adults said they had little or no confidence in American companies to develop and use AI responsibly, while 67% had little or no confidence in the U.S. government to regulate AI effectively. Around seven in ten also expected AI to make their personal information less secure. [5]
That creates an uncomfortable situation. Businesses are giving AI more capability and more authority at the same time that much of the public remains skeptical of the technology, skeptical of the companies deploying it, and skeptical that government regulation will solve the problem for them.
Slowing AI adoption until everyone is comfortable with the technology doesn't seem practical given the scale of investment and adoption. However, I do think there is a path forward, we just have to ask the right questions like…
What can businesses do to earn that trust?
AI governance has mostly focused on the system
There is a lot of quality work happening in AI governance today by really smart people. Organizations are working on identity, permissions, model benchmarking, monitoring, logging, access controls, policy enforcement, and increasingly “Know Your Agent” frameworks. These controls help answer important questions like… Who is this agent? What is it allowed to do? What systems can it access? How is it behaving? But there is another, more fundamental layer that I think is getting less attention.
What happens to the output?
An AI system generates a report. An agent makes a recommendation. An API returns a result. A chatbot gives a customer an answer. That information gets copied, downloaded, forwarded, stored, or passed into another system. Once it leaves the environment that produced it, much of the governance context gets left behind and the questions become different…Who or what created this? Has it been modified? When was it created? Was the system authorized to produce it? And more importantly, Should I still trust it? That is the accountability gap I think we need to address.
Trust shouldn't be permanent
AI makes this even more important because AI systems are unusually dynamic. Models change. Prompts change. Agents gain and lose capabilities. Permissions get updated. Systems get compromised. Organizations change vendors. A trust decision made six months ago may no longer represent the system operating today. Yet many approaches to digital trust are built around proving that something was valid at one point in time, but for AI, I think we need to go further. We need to be able to establish provenance and integrity, but we also need to be able to manage the lifecycle of trust itself. Trust should be able to expire, be renewed, or when circumstances change, be revoked.
Models change. Agents change. Permissions change. Trust should change too.
The ecosystem is already moving in this direction and there are plenty of signs that output-level accountability is becoming more important.
The EU AI Act's Article 50 includes transparency requirements for certain AI-generated content. Specifically, it requires providers of systems generating synthetic text, audio, image, or video to make outputs machine-readable and detectable as artificially generated or manipulated, subject to the law's conditions and exceptions. [6]
C2PA is building technical standards around digital content provenance and authenticity, including cryptographically verifiable information about a piece of content's source and history. [7]
We're also seeing AI vendors experiment with watermarking, while identity and governance vendors increasingly focus on knowing which AI agents are operating inside an organization. I don't see these approaches as mutually exclusive, but actually point toward the same broader conclusion.
AI governance needs to become more technical, more verifiable, and more portable.
Policy matters. Monitoring matters. Identity matters. But eventually, the information AI produces also needs a way to carry accountability with it.
Where BlkSeal fits
That is the problem we're working on with BlkSeal. BlkSeal is a content integrity platform designed to let people, applications, and AI systems sign digital outputs when they're created, share them through normal workflows, and verify them later when trust matters. The initial focus is AI-generated content and high-consequence digital workflows, but the underlying idea is broader. Text, API responses, reports, documents, media, and other digital artifacts can all benefit from being able to establish where they came from and whether they've changed.
BlkSeal is powered by BlkBolt, our proprietary machine-learning-based encoding technology. It gives us a different approach to signing without requiring every use case to inherit the full certificate and key-management model of traditional PKI. I'm not arguing that PKI, C2PA, watermarking, or existing governance tools are wrong. They each solve useful problems, but I think there is simply a practical space between doing nothing and deploying heavyweight trust infrastructure into every workflow. That's the space we're building for.
With BlkSeal, the goal is to make trust simple, portable, and manageable.
- Sign content when it's created.
- Share it normally.
- Verify it when trust matters.
- Manage that trust when circumstances change.