COMMENTARY: The launch of Meta’s Muse on Sept. 8, 2026 brought personal AI agents closer to the mass market. Muse can act through software on a user’s behalf, from sending emails to booking travel. But the significance goes far beyond Meta.
For decades, the internet maintained a practical distinction between humans and machines. Software used APIs; people used browsers, accounts, passwords, and biometrics. The separation was never absolute, but digital systems could often assume that an action through a human interface came from a person.
[SC Media Perspectives columns are written by a trusted community of SC Media cybersecurity subject matter experts. Read more Perspectives here.]
AI agents challenge that assumption. They use the same browsers, complete forms, conduct conversations, and perform actions once taken as signs of human presence. Generative AI erodes other signals: text does not prove human authorship, a voice does not prove someone is speaking, and a face on camera does not guarantee a person is present.
Humans and machines have not become identical. The environment is losing its ability to tell them apart.
The "Frankensteinization of the Internet"
I call this the "Frankensteinization of the Internet" — different classes of actors sharing a digital environment whose ability to distinguish them is disappearing.
The metaphor has two meanings. First, a new whole emerges from heterogeneous parts. Humans remain humans, software remains software, and agents remain agents. Their nature has not changed; our ability to recognize them has.
Related reading:
Second, we created this environment ourselves. We designed it around human ways of reading, speaking, seeing, and acting, representing people through accounts, devices, faces, voices, and behavior. Then we taught machines to reproduce those manifestations.
The digital environment created by humans is gradually losing the ability to recognize its own creator.
Who are we trying to identify?
The natural response is better machine identification. Verifiable agent identity can establish an agent’s credentials, permissions, and relationship to those it represents. Personhood credentials seek to let individuals prove they are human without necessarily disclosing who they are.
Both approaches matter. But we cannot predict every class of digital actor that will exist in five or 10 years. If our strategy depends on identifying every new kind of machine, the problem will continually change with technology.
Perhaps we should reverse the question. Instead of endlessly proving that something is not AI, we need a reliable way to establish something positive: This is human.
The challenge is not simply that machines are becoming indistinguishable from humans. Humans are becoming indistinguishable as humans in digital environments. For digital identity, this changes the point of reference.
Do we always need the distinction?
Not necessarily. If my AI agent buys an airline ticket within the authority I have granted it, the airline may not care who completed the individual steps.
Other environments may choose differently. A community might allow agents if they disclose themselves. A financial service might require a human for certain actions. Another platform might deliberately treat humans and machines alike.
But meaningful choice requires the technical ability to distinguish participants. Without it, a decision not to distinguish is no longer a policy choice. It is simply the only available option.
We can choose to regulate, restrict, or permit only what we are capable of distinguishing.
Human distinguishability does not mean excluding AI. It makes it possible to decide how humans and machines should coexist.
Back to the human
For years, digital identity has focused on one question: “Who are you?” Passwords, tokens, devices, and biometrics helped connect actions to identities. Whether the actor was human was often assumed.
Now another question comes first: “Is it human?”
Answering it does not require identifying every possible machine or granting humans special privileges. It means restoring our ability to positively recognize human participation in digital environments.
Perhaps this is becoming a foundational requirement of digital identity in the age of AI agents. And if Muse helped make the problem visible to millions, then for that, at least, we can say: thanks, Meta.
