The Problem with Knowing a Name
For decades, digital security rested on a simple premise: if you can verify who someone is, you can decide whether to trust them.
A username. A password. A one-time code. A badge scan. These static identity markers were the gatekeepers of our systems — and for a while, they worked well enough.
They no longer do.
Today’s threat landscape has fundamentally changed. Credentials are stolen, shared, phished, and replicated on an industrial scale. Insider threats don’t announce themselves. AI agents operate autonomously across enterprise systems, impersonating legitimate users with precision. And the old model — verify once at the front door, then trust indefinitely — has become a critical vulnerability hiding in plain sight.
The question security leaders must now ask is not “who is this?” but “is this person or agent behaving the way we’d expect them to, right now, in this context?”
That is the shift at the heart of intent-based AI security, and it represents one of the most important evolutions in identity and access management of the past decade.
What Is Static Identity and Why Has It Failed?
Static identity verification is exactly what it sounds like: a fixed check against a known credential or identifier. A password matches a stored hash. A token is valid. An email address is confirmed. Access is granted.
The problem is that static identity answers only one question: does this credential exist? In many cases it says nothing about:
- Whether the person using the credential is who they claim to be
- Whether their behaviour is consistent with their historical patterns
- Whether the request itself makes sense given the time, location, device, or workflow context
- Whether an AI agent acting on a user’s behalf is operating within expected parameters and with the appropriate authorisation.
In a world where credential theft is routine and AI-driven impersonation is increasingly sophisticated; a static ID check is the digital equivalent of checking a library card before letting someone into the server room. It confirms a token. It cannot confirm intent.
The Rise of Context Signals: A New Foundation for Identity Trust
If static identity verification is a single data point — a yes or no — context signal aggregation is a rich, multi-dimensional portrait of who is acting, how and why.
Context signals are pieces of behavioural and environmental data drawn from trusted sources that, when stitched together, paint a far more reliable picture of identity at the moment of action. They include:
Behavioural signals: How does this user typically navigate the system? What data do they usually access, at what times, and in what volume? Is this request consistent with their established patterns?
Device and network signals: Is this the device they normally use? Is the network familiar? Has the geolocation shifted unexpectedly?
Session signals: How long have they been active? What actions have they already taken in this session? Does the sequence of steps make sense?
Workflow and role signals: Does this request align with what someone in this role would legitimately need to do? Is it consistent with the task they were assigned?
Temporal signals: Is this action being taken at an unusual time? Does the cadence of actions suggest automation, duress, or anomaly?
No single signal is definitive. But when multiple trusted sources are cross-referenced and stitched together in real time, they create a probabilistic identity confidence score that is far more robust than any credential check alone.
This is the foundation of the next generation of identity security — and it requires infrastructure capable of integrating, correlating, and interpreting these signals continuously, without introducing friction for legitimate users.
Intent-Based AI Security: Shifting from Authentication to Understanding
Intent-based AI security takes the concept of context signal aggregation further, applying machine learning and behavioural analytics to understand not just who is acting, but what they are trying to accomplish — and whether that intent is consistent with legitimate use.
Rather than asking “is this credential valid?”, intent-based security asks:
- Is this action consistent with the user’s historical intent patterns?
- Does the request sequence suggest a legitimate workflow or a reconnaissance attempt?
- Is the volume, velocity, or directionality of data access anomalous?
- Are multiple low-risk signals combining into a high-risk behavioural profile?
This approach fundamentally changes where security detection happens. Traditional models are perimeter-based: trust is established at the boundary and assumed to persist. Intent-based AI security is continuous: trust is earned moment to moment, with each action evaluated against a dynamic model of expected behaviour.
The practical implication is profound. A legitimate employee who suddenly begins exfiltrating large volumes of sensitive documents will trigger anomaly detection, even if their credentials are entirely valid. An attacker using stolen credentials who doesn’t know the victim’s usual workflow will reveal themselves through behavioural inconsistency, long before they reach a critical asset.
The Agent Problem: Why Human Identity Is No Longer Enough
Here is where modern security faces an entirely new frontier.
AI agents (autonomous software entities that can log in, execute tasks, query databases, send communications, and interact with systems on behalf of user) are now a mainstream feature of enterprise environments. They operate in CRMs, financial systems, HR platforms, customer service tools, and beyond.
And they have identities too.
The question of who is acting? is no longer limited to human users. It must now encompass:
- AI agents operating autonomously, potentially with access credentials assigned to a human user
- Orchestration layers that chain multiple AI models together in complex workflows
- AI-to-AI interactions, where one agent instructs another without any direct human oversight
Traditional identity frameworks were simply not designed for this. They assume a human at the keyboard. They verify a static credential. They cannot assess whether an AI agent’s behaviour is within expected parameters, or whether it has been manipulated, hijacked, or is acting on a malicious instruction that originated outside the organisation’s control.
Intent-based AI security addresses this by applying the same continuous behavioural analysis to agent activity that it then applies to human activity. Expected agent behaviour — the tasks it performs, the data it touches, the systems it calls — is modelled as a baseline. Deviations from that baseline trigger scrutiny, regardless of whether a valid credential is presented.
This is not a theoretical concern. As agentic AI systems proliferate across enterprise environments, the attack surface they represent will become a primary vector for sophisticated threats. Securing that surface requires security frameworks that understand agent intent, not just agent identity.
How Sedicii Approaches the Context Signal Challenge
At Sedicii, our approach to identity has always centred on the principle that trust must be earned through verified evidence from independent, trusted sources, not assumed only from credentials.
Our zero knowledge proof technology was designed from the outset to enable identity verification without unnecessary data exposure — proving something is true without revealing the underlying data. That same privacy-preserving philosophy extends to how we think about context signals and behavioural identity.
The challenge in building robust context signal infrastructure is not just technical. It is also deeply tied to data governance, privacy regulation, and organisational trust. Aggregating behavioural signals from multiple sources raises important questions about consent, data minimisation, and the rights of individuals whose behaviour is being continuously assessed for risks.
Sedicii’s approach navigates these tensions by:
Anchoring signals in independent trusted sources. Not all data is equal. Context signals drawn from verified, high-integrity sources carry more weight than those from unvalidated inputs. Our framework prioritises signals that can be cryptographically attested or sourced from authoritative systems of record.
Stitching without storing unnecessarily. The goal is not to build a permanent surveillance profile but to enable real-time risk assessment. Privacy-preserving computation techniques allow signals to be evaluated without requiring centralised storage of sensitive behavioural data.
Applying proportionate scrutiny. Low-risk activities in familiar contexts require minimal verification. High-risk actions — particularly those involving sensitive data, privileged access, or unusual behavioural patterns — trigger elevated assurance requirements. This proportionality reduces friction for legitimate users while concentrating scrutiny where it matters most.
Extending the framework to AI agents. As agentic AI becomes a standard feature of enterprise operations, identity and access management frameworks must evolve to treat agents as first-class entities with their own behavioural baselines, access policies, and intent models.
From Point-in-Time Verification to Continuous Trust
The shift from static identity to contextual, intent-based security is, at its core, a shift in how we think about trust.
Static ID asks: did this person prove they are who they say they are at the moment they logged in?
Continuous contextual security asks: does this person or agent continue to behave in a way that is consistent with who they claim to be, across every action they take, over time?
The answer to the first question can be obtained with a stolen password. The answer to the second requires authentic behaviour — something that is far harder to fake at scale, over time, and across complex workflows.
This is why intent-based security is not simply a more sophisticated version of multi-factor authentication. It is a fundamentally different model of trust — one that reflects the realities of modern threats, modern users, and the increasingly autonomous systems that operate alongside them.
The Path Forward: Building Identity Infrastructure Fit for the AI Era
For security leaders and identity architects, the implications of this shift are clear:
Static credentials must be supplemented with continuous behavioural context. Point-in-time verification is a necessary but insufficient condition for access security. Organisations need infrastructure that aggregates context signals in real time and adjusts trust levels dynamically.
AI agents require identity frameworks of their own. The assumption that identity security is a human problem is no longer valid. Agentic systems need behavioural baselines, access governance, and continuous monitoring just as human users do.
Privacy and security are not in opposition. Effective context signal aggregation can be achieved through privacy-preserving approaches that evaluate risk without unnecessary exposure of underlying data. These approaches are not only technically feasible — in an era of increasing regulatory scrutiny, they are commercially essential.
Intent matters as much as identity. The goal of modern security is not just to verify a credential. It is to establish confidence that the entity requesting access, human or agent, is authorised and doing so for legitimate purposes, in an expected context, in a manner consistent with their established patterns.
Sedicii is building the infrastructure to make this possible: identity security that goes beyond the name on the door, to understand the intent behind the action.
Conclusion: Identity Without Context Is Incomplete
The era of static identity verification is ending — not because passwords were a bad idea, but because the world they were designed to protect is disappearing.
Today’s enterprise environments are distributed, multi-cloud, heavily automated, and increasingly populated by AI agents operating alongside human users. The threats they face are sophisticated, adaptive, and explicitly designed to circumvent credential-based security.
Meeting that challenge requires a new model: one built on context signals from trusted sources, stitched together in real time, and interpreted through the lens of expected intent. One that applies equally to the human employee, the third-party contractor, and the AI agent operating on their behalf.
Static ID told you a name. Context-driven, intent-based security tells you the truth.
Sedicii builds privacy-preserving identity infrastructure for organisations that need to verify trust with other organisations without compromising sensitive data, using protocols like zero knowledge proofs and multiparty computations. To learn more about our approach to context signal aggregation and intent-based identity security, get in touch with our team at contactus@sedicii.com.