The photo used to be the proof

For as long as insurers have processed claims, the photograph has done a specific job: it stood in for the event itself. A cracked windscreen, a flooded kitchen, a dented bumper. The image was submitted, a handler looked at it, and in most cases what they saw was treated as a reasonably faithful record of what happened. Claims teams built entire triage models around that assumption. Photo evidence looked consistent with the claim narrative, so the claim moved forward. Photo evidence looked odd, so it got a second look.

That assumption is now the weak point in the process. Generative AI tools can produce a photorealistic image of storm damage, a flooded room, or a written repair estimate that never happened in the physical world, and they can do it without leaving the kind of obvious tells that used to give fraud away. A poorly cloned dent, a mismatched shadow, a receipt with an implausible font. Those tells are becoming rarer, not because fraud is harder to commit, but because the tools generating the evidence have improved. The danger isn’t that synthetic claims evidence looks flawless. It’s that it looks completely ordinary.

The problem is already showing up in claims departments

Major Japanese insurers Mitsui Sumitomo and Aioi Nissay Dowa have introduced AI image detection tools specifically to identify manipulated claims documents, a direct response to what Asia Insurance Review reports as a growing risk of fraud involving generative AI in claims evidence. When two large carriers move to address the same problem at the same time, it’s a reasonable signal that the risk has moved from theoretical to operational.

How synthetic evidence enters the claims workflow

The exposure sits at first notification of loss and through the claims documentation stage, wherever a claimant is asked to upload photographic or documentary evidence without a loss adjuster physically present. That covers a large and growing share of claims, particularly low-value motor, property and contents claims that are increasingly processed through self-service digital channels precisely because they’re low value and high volume.

What gets submitted typically falls into a few categories: photographs of physical damage, photographs of the item or property before the loss for comparison, repair estimates and invoices from third-party contractors, and sometimes correspondence or messages supporting the claim narrative. A claims handler reviewing this evidence is checking for internal consistency. Does the damage match the stated cause of loss? Does the invoice come from a plausible supplier at a plausible price? Does the image look like it was taken in the environment the claimant describes?

That review process was designed around human-generated evidence, where fabrication left traces. A staged photo had lighting inconsistencies. A doctored invoice had font mismatches or alignment errors. A reused image from a previous claim or a stock photo could sometimes be caught on a second look, or through basic reverse image search.

Generative image tools remove most of those traces. A model can produce damage that never occurred, in a setting that plausibly matches the claimant’s description, without the compositing errors that used to flag manipulation. It can also generate documents, invoices and repair quotes with correct formatting and internally consistent detail. The handler is applying exactly the same contextual judgement they always have. The evidence has simply stopped behaving the way evidence used to behave.

What this does to the claims desk

The operational pressure lands in two directions at once. Claims volumes in digital-first channels are high, and the commercial incentive is to process straightforward claims quickly. Adding manual scrutiny to every submission is not viable; it slows genuine claimants, adds cost, and doesn’t scale. But treating every image as trustworthy by default means a growing category of synthetic fraud passes through unchallenged, because the visual and contextual cues handlers were trained to notice are no longer reliable indicators.

The false-negative risk is the one that matters most here. A handler who has never been shown a convincing AI-generated damage photo has no internal benchmark for what one looks like, because until recently there wasn’t one. That’s not a competence gap. Claims handlers are skilled at judging narrative plausibility, at spotting behavioural inconsistency, at applying institutional knowledge about typical loss patterns. They were never equipped, and never needed to be, to detect pixel-level generation artefacts or image reuse across submissions. That’s a different kind of pattern recognition entirely, and it sits outside what contextual review can reasonably be expected to catch.

Applying scrutiny where it’s needed, not everywhere

This is where an authenticity layer changes the shape of the problem, rather than trying to replace the handler’s judgement. Humanly analyses submitted images and documents for signals that may indicate synthetic generation or manipulation, such as indicators of AI generation or inconsistencies that don’t align with a genuine capture, and surfaces those signals alongside the claim inside the existing review workflow.

The result is selective confidence: scrutiny gets applied proportionally rather than uniformly. A claim with no authenticity indicators can move through the normal fast path. A claim carrying signals that may indicate manipulation gets flagged for a closer look by the handler, who then applies exactly the contextual judgement they’re good at, informed by information they wouldn’t otherwise have had. Humanly doesn’t approve, reject or automate anything. It gives the human reviewer a reason to look twice, and the evidence to know where to look. Human first, AI guarded.

As claims evidence becomes easier to fabricate convincingly, the value isn’t in replacing the adjuster’s judgement. It’s in restoring the reliability of what they’re judging.

Talk to us

If synthetic fraud is becoming a live concern in your claims evidence intake, speak to our sales team about how Humanly fits inside your existing review workflow.

The proxy that stopped being reliable

For decades, a photograph, a screenshot or a signed document did useful work in a courtroom without anyone having to argue about it. It stood in for the event it depicted. A text message thread was treated as a record of what was said. A photo of a car after a collision was treated as a record of the damage. Courts built entire procedural habits around the assumption that digital evidence was, in the ordinary case, a faithful trace of something that actually happened.

That assumption is no longer safe. Generative AI tools can now produce a convincing screenshot of a text conversation, a photograph of an injury that never occurred, or a document that looks correctly formatted and internally consistent, in minutes and without specialist skill. The evidence doesn’t need to be perfect to do damage. It only needs to be ordinary enough to pass the level of scrutiny it would normally receive, which for most exhibits is not very much at all.

The alarm has already been raised

The National Center for State Courts has warned that courts currently lack standardised protocols to detect or evaluate AI-generated or AI-manipulated evidence, a gap it describes as a systemic risk to public trust in the judicial process, as reported by The Florida Bar. That warning is not speculative. It is a description of a gap that exists right now, in active caseloads, across civil and criminal matters alike.

Where the check breaks down

Digital evidence enters a case file at several points: attached to a complaint, produced in discovery, appended to a motion, or introduced as an exhibit at trial. It typically arrives as a photograph, a scanned or exported document, or a screenshot of a message, email or account statement. Somewhere in the chain, a paralegal, clerk, opposing counsel or judge looks at it and asks the questions courts have always asked: does this look consistent with the rest of the record, is the format right, does the story it tells hold together.

That is contextual reasoning, and legal professionals are good at it. It is not, however, the same skill as spotting a pixel-level artefact, an inconsistent shadow, a font kerning error introduced by a generation pipeline, or a reuse pattern that shows the same synthetic template appearing across unrelated filings. Nobody trained lawyers or judges to do that, because until recently there was no need. The evidentiary system was built to test credibility and consistency, not to forensically examine whether a JPEG was ever captured by a camera at all.

This is where the traditional check quietly stops working. A document can be internally consistent, plausible in context, and still be synthetic in origin. Metadata, long treated as a reliable authenticity signal, can be stripped, edited or fabricated by the same tools that generate the content. Asking a party whether they used forensic tools to verify their own submission, as some proposed bench cards now suggest, is a reasonable question but not a technical answer. The gap is not a failure of legal skill. It is a mismatch between what reviewers were trained to look for and what synthetic fraud actually looks like.

The operational cost of unresolved doubt

The practical consequence lands on court staff and litigators long before it reaches a judge’s bench. Every exhibit, screenshot and scanned document is now a candidate for challenge, and challenging one properly means expert declarations, forensic retainers and delay, all in a system already under docket pressure. Escalate every questionable submission and the caseload grinds to a halt. Escalate none and the door stays open to the exact risk the NCSC is describing: fabricated evidence admitted because it looked ordinary, or genuine evidence dismissed because a party successfully cast doubt on it.

That second failure mode, sometimes called the Liar’s Dividend, deserves as much attention as the first. Once everyone knows convincing fakes are possible, a party can cast doubt on authentic evidence simply by suggesting it might be synthetic, with no need to prove anything. The absence of a reliable way to check authenticity doesn’t just let fabrications through. It corrodes confidence in evidence that is entirely genuine, which is arguably the more corrosive long-term effect on evidence integrity.

What an authenticity layer changes

The answer is not asking clerks, paralegals or judges to become forensic examiners, and it is not automating admissibility decisions. It is giving reviewers explainable authenticity signals on the images and documents in front of them, so that scrutiny can be applied proportionally rather than uniformly. A submission with no unusual indicators can move through review at its normal pace. One flagged with signals that may indicate manipulation, such as inconsistencies in compression patterns, structural anomalies inconsistent with the claimed capture method, or reuse markers linking it to other submissions, can be routed for the closer, more expensive scrutiny it actually warrants.

This is the principle behind an authenticity layer: human first, AI guarded. The reviewer still makes the judgment call about credibility, context and admissibility. What changes is that they are no longer relying solely on contextual reasoning to catch something that was never designed to be visible to contextual reasoning in the first place. Humanly analyses images, documents, screenshots and written statements, and returns indicators that may warrant closer scrutiny, not verdicts. The decision, as it should be, stays with the person qualified to make it.

For litigation support teams, court administrators or legal ops functions weighing up how to handle this exposure, the conversation worth having is about where selective confidence fits into an existing review workflow, not about replacing it.

Talk to Humanly’s team about applying selective confidence to evidence review.

The document used to be the proof

For as long as referencing has existed, a payslip meant something. A bank statement, a reference letter, a passport scan: these were treated as reliable proxies for the facts they represented. If the numbers matched and the format looked right, the underlying reality was assumed to match too. Letting agents built entire referencing workflows on that assumption, and for a long time it held.

It no longer holds, because the cost of producing a convincing fake has collapsed. Generating a payslip that carries the right typography, the right employer logo, the right National Insurance format and a plausible pay history now takes minutes and requires no forgery skill at all. The same is true of reference letters, tenancy histories and identity documents. The tenant referencing sector is not facing more forgery. It is facing a different kind of evidence problem, where the fake is not a poor copy of the real thing but a synthetic original with no real thing behind it at all.

The scale is no longer anecdotal

This is not a theoretical risk. Goodlord’s analysis of over a million tenant references found that suspected fraud rose by almost 40% during 2025, with fake employment references up 226.6% year-on-year and referee fraud and identity manipulation both surging as fraudsters target the verification process itself rather than just the paperwork, as reported by Letting Agent Today.

Where the checks were built for a different problem

Tenant referencing typically runs at a single point in the process: application submission. A prospective tenant uploads payslips or bank statements to evidence income, a reference from a current or previous landlord, sometimes a letter from an employer, and identity documents to satisfy Right to Rent checks. A referencing analyst or letting agent reviews this bundle, cross-checking figures against declared income, checking that names and addresses are internally consistent, and occasionally placing a call to a referee.

This review process was designed to catch human error and clumsy forgery: a payslip with the wrong tax year, a reference letter with an inconsistent job title, numbers that don’t add up. It was never designed to catch a document that is internally consistent because it was generated to be internally consistent. AI-generated payslips don’t have typos in the tax code because the model wasn’t copying a real payslip badly, it was generating a plausible one from scratch. The same applies to reference letters and, increasingly, to entire fabricated identities built from a combination of a bogus employer, a doctored ID and an invented referee who will confirm whatever the applicant needs confirmed.

This is the shift that matters for anyone running a referencing team: the fraud has moved from the document to the identity behind it. A single forged payslip is one point of failure to check. A coherent fake identity, with a matching employer, a plausible referee and consistent paperwork across every touchpoint, is a system built to pass every check an agent has time to run. Reviewers are good at spotting inconsistency and implausibility. They are not equipped to spot pixel-level generation artefacts, font irregularities that don’t register consciously, or the reuse of a document template across multiple unrelated applications, because none of that was ever part of the job description.

What this costs the referencing desk

The operational pressure this creates is straightforward but unforgiving. Referencing volumes have not gone down, and every application still needs a decision within a commercially sensible timeframe. Agents cannot escalate every application for manual deep-checking without breaking turnaround times and frustrating genuine tenants, most of whom are exactly who they say they are. But treating every application as low risk means the fabricated ones, the ones built specifically to look ordinary rather than suspicious, pass straight through.

That is the real danger here: synthetic evidence is not designed to look perfect, it is designed to look unremarkable. A convincing fake payslip doesn’t draw attention to itself. It sits in the pile with forty other legitimate ones and gets the same glance. The false-negative risk isn’t a rare edge case, it’s the default outcome of a system with finite review time and an applicant population that increasingly includes people using the same accessible tools. And the data suggests fraud is not evenly distributed either, with certain regions and higher-value properties carrying meaningfully higher exposure, which means uniform scrutiny is both inefficient and misdirected.

Applying scrutiny where it’s actually needed

This is where an authenticity layer changes the shape of the problem rather than the workload. Humanly analyses the documents and images submitted as part of a referencing application, whether that’s a payslip, a bank statement, a reference letter or an identity document, and surfaces explainable signals that may indicate AI generation or manipulation, directly inside the review workflow the agent already uses.

The principle is selective confidence: scrutiny applied proportionally, not uniformly. Instead of every application getting the same cursory check or every application triggering a manual escalation, the referencing team gets a signal that tells them where a closer look is warranted and, just as importantly, where it isn’t. This is decision support, not decision automation. Humanly does not approve or reject a tenancy application; it gives the human reviewer the same kind of contextual judgement they already bring to a case, plus visibility into the kind of artefact they were never trained to see. Human first, AI guarded.

For a sector where document authenticity used to be assumed and is now the actual point of failure, that additional layer of evidence integrity is not a nice-to-have. It is what lets a referencing team keep moving at volume without quietly waving fabricated identities through.

Talk to us

If your referencing workflow needs a way to spot AI-generated or manipulated documents without slowing down every genuine application, speak to our sales team about adding an authenticity layer to your existing process.

The question we hear most often

“How does AI fit into decision workflows?”
“What happens if it makes an incorrect judgement?”

These questions come up in almost every conversation with operations leaders, risk teams and decision owners. They are valid concerns. They reflect accountability, governance and the need for control.

But they are based on an assumption that no longer reflects how modern decision environments actually need to work.  The rules have changed – the game has changed.

Evidence is no longer simply provided, it can be generated, edited or enhanced using AI. As a result, assessing authenticity is becoming very difficult, through manual human-review alone.

At Humanly, we think business leaders need to reframe the question

The question we need to ask 

How effective will our decision making be without the use of AI?

This reframing shifts the focus from AI risk to decision integrity.

In many operational workflows, decisions are only as reliable as the evidence they are based on. If the nature of that evidence is changing, then the way it is evaluated needs to change as well.

This is where AI becomes relevant, not as a replacement for human judgement, but as a way to strengthen how inputs are interpreted before decisions are made.

The real shift: evidence is becoming harder to interpret

Across industries, organisations rely on user-submitted inputs:

• Claims and supporting documentation
• Identity records and verification materials
• Property images and survey evidence
• Healthcare documentation and eligibility data

Historically, these inputs were assumed to be either genuine or obviously problematic.

That assumption is becoming less reliable.

Content can now be created or altered in ways that appear credible at first glance. This does not mean every submission is untrustworthy. It does mean that authenticity is no longer always visible through inspection alone.

As a result, decision-making increasingly depends on:

• The ability to assess authenticity signals
• The consistency of that assessment across reviewers
• The context available at the moment a decision is made

Where human-only decision workflows can struggle

Human expertise remains central to decision-making. However, when evaluating potentially synthetic or manipulated inputs, certain limitations can emerge:

• Inconsistent interpretation – Different reviewers may reach different conclusions when signals are ambiguous.
• Limited visibility – Some indicators of generated or altered content are not easily detectable without additional analysis.
• Time constraints – High-volume workflows often limit the depth of manual investigation.
• Expanding input types – The variety and complexity of submissions continue to increase.

These challenges are structural and reflect a shift in the nature of evidence, not a lack of reviewer capability.

Humanly’s role: an Authenticity Intelligence layer within the workflow

Humanly is designed to sit inside decision workflows as a source of structured perspective on authenticity.

It does not replace existing systems or human reviewers. Instead, it introduces a consistent way to answer a question that is often handled informally:

“Can we trust this input?”

Humany provides explainable insight and perspective into decision workflows.

A typical workflow incorporating Humanly looks like this:

1. Evidence is submitted by a user (documents, images, data inputs)
2. Humanly analyses the submission for authenticity signals
3. Signals are surfaced within the workflow
4. A human reviewer evaluates both the submission and the signals
5. The organisation makes the final decision

This approach preserves accountability while improving the quality of context available to decision-makers.

What Humanly contributes

Humanly is designed to support decision workflows by:

• Analysing authenticity signals across submitted content
• Providing structured outputs to support reviewer interpretation
• Improving consistency in how authenticity is assessed
• Integrating into existing review environments

This can help teams move beyond purely visual or intuition-based assessments, particularly in edge cases.

What Humanly does not do

To be clear:

• Humanly does not make approval or rejection decisions
• It does not replace underwriting, claims or eligibility logic
• It does not remove human accountability

Its role is to inform decisions, not to make them.

Reframing the risk: the issue is not “AI making mistakes”

A common concern is that AI could make incorrect decisions.

In workflows where AI is autonomous, that concern is valid. However, in many real-world decision environments, the more immediate risk is different:

• Decisions being made on inputs that have not been sufficiently evaluated for authenticity
• Over-reliance on manual review in contexts where inputs are becoming harder to interpret

From this perspective, AI is not the source of the problem. It is part of how organisations can better understand and manage that uncertainty.

Industry applications: where this matters in practice

Insurance: claims decisions depend on evidence quality

Insurance workflows rely heavily on submitted evidence such as images, documents and written descriptions.

Humanly can support claims teams by:

• Providing additional context on the authenticity of submitted materials such as images of scenes, vehicle images, damage assessments, weather conditions, scene manipulation
• Thus helping identify cases that may require further review
• Supporting consistency across high-volume claims environments

You can read more here: Insurance AI fraud detection

Healthcare: documentation-driven processes require trust

Healthcare systems process large volumes of documentation across claims, eligibility and administration.

Humanly is designed to:

• Analyse images for prescription claims, such as weight variances or injury
• Support teams managing complex, high-throughput workflows
• Provide additional context where submissions are uncertain

This helps maintain confidence in inputs without interfering with clinical judgement.

You can read more here: Healthcare AI document verification

Identity: authenticity beyond traditional verification

Identity workflows increasingly rely on digital, user-submitted content.

Humanly can complement existing systems by:

• Assessing whether submitted IDs may be synthetic or altered. Such as drivers licences, passports or any form of identity documents
• Supporting reviewers in ambiguous or edge-case scenarios
• Adding an authenticity-focused layer alongside identity verification processes

You can read more here: Identity authenticity intelligence

Property & Retrofit: distributed evidence creates new challenges

Property and retrofit workflows often depend on images, surveys and third-party submissions collected remotely.

Humanly can help:

• Analyse government grant claims such as ECO4 or Warm Homes Plan grants. 
• Identify facilities management claims with trade error
• Health and safety claims – scene manipulation
• Support teams operating across distributed environments
• Provide context when authenticity is not immediately clear

You can read more here: Property & retrofit verification

Why Humanly is best understood as decision support

Humanly is not a decision engine. It is a decision support layer focused on authenticity.

This distinction is important:

• Humans provide judgement, accountability and context
• Humanly provides structured analysis of authenticity signals
• Decisions become more informed without becoming automated

This model aligns with how many organisations are adapting to AI, by augmenting human capability rather than replacing it.

It’s more about Building decision confidence in a synthetic world

As AI-generated and AI-manipulated content becomes more accessible, organisations face a practical challenge:

How do we maintain confidence in decisions when the inputs themselves may be uncertain?

Humanly is designed to address this by:

• Introducing structured authenticity analysis into workflows
• Supporting human reviewers with additional context
• Improving consistency and clarity in decision-making

The goal is not to eliminate uncertainty. It is to ensure that decisions are made with a clearer understanding of the inputs behind them.

If your organisation relies on reviewing user-submitted evidence, the next step is to understand how authenticity intelligence can support your workflow.

Speak to our enterprise sales team.

Organisations operating in regulated environments often approach AI cautiously. Concerns around governance, accountability and regulatory compliance are legitimate. Yet the reality of modern operations is that many decision workflows now involve digital evidence that may be AI-generated or AI-manipulated.

In this environment, relying solely on manual human review can introduce its own regulatory risk. When synthetic or manipulated submissions become difficult to detect with the human eye alone, organisations may face increased exposure to errors in claims assessment, identity verification or evidence review.

The emerging question for regulated industries is no longer simply “Is AI safe to deploy?” but increasingly “What risks arise when AI is not used to support human judgement?”

Humanly’s approach is to introduce AI-assisted authenticity analysis that supports human reviewers, helping organisations assess digital submissions more reliably while keeping people firmly in control of the final decision. This human-first, AI-guarded model is designed to strengthen decision integrity rather than automate it.

The Regulatory Challenge of Operating in an AI-Generated World

Generative AI tools have made it significantly easier to create or modify digital content. Images, documents and other forms of digital evidence can now be produced with increasing realism.

For organisations that rely on human-submitted evidence — such as insurance claims, benefits applications, healthcare documentation or financial verification — this creates a new operational challenge.

• Human reviewers may still be responsible for assessing:
• Photographic evidence
• Identity documentation
• Medical or health-related submissions
• Supporting documentation for claims or benefits

However, the underlying information environment has changed. Some submissions may now contain synthetic or AI-manipulated elements that are difficult to identify through visual inspection alone.

When review workflows depend entirely on manual judgement, the risk is not only fraud exposure. There can also be downstream regulatory implications, including:

• Inaccurate claim approvals or rejections
• Consumer duty concerns if decisions are made on manipulated evidence
• Patient safety risks if health-related documentation is misinterpreted
• Operational governance issues if authenticity checks are inconsistent

In other words, the shift to a synthetic content environment can introduce new forms of regulatory exposure if authenticity controls do not evolve alongside the technology landscape.

When Manual Review Alone Becomes a Regulatory Risk

Human expertise remains central to decision-making in regulated sectors. Experienced reviewers bring contextual judgement, domain knowledge and accountability that automated systems cannot replicate.

However, manual review was designed for a world where most evidence was assumed to be naturally produced.

When generative AI tools can alter or generate evidence with high realism, the limitations of visual inspection become more visible. Some organisations have observed that manipulated submissions can be difficult for reviewers to detect reliably without technical assistance.

Consider a hypothetical healthcare-related scenario.

Example: Synthetic Evidence in Pharmacy Submissions

Imagine a workflow where patients submit photographic evidence relating to prescription eligibility or medication access. A person may attempt to alter physical appearance in submitted images or use generative tools to modify digital evidence.

From a reviewer’s perspective, the image may appear plausible. Without specialised analysis tools, detecting subtle manipulation could be extremely challenging.

If such submissions are incorrectly accepted or rejected, the consequences may extend beyond operational inefficiency. They may influence:

• Medication eligibility decisions
• Patient treatment pathways
• Regulatory reporting accuracy
• Consumer fairness obligations

In these contexts, the question becomes less about whether AI introduces risk and more about whether organisations have adequate controls to assess AI-influenced evidence.

A Different Way to Think About AI in Regulated Workflows

Many discussions about AI adoption frame the technology as a replacement for human decision-making. That framing understandably raises concerns among compliance teams and regulators.

Humanly’s view is different.

AI should not replace the human reviewer. Instead, it can introduce an additional analytical perspective into the decision process.

Rather than automating approvals or rejections, Humanly is designed to:

• Analyse submissions for authenticity signals associated with AI generation or manipulation
• Surface insights that may warrant additional human attention
• Provide decision-support context to reviewers
• Allow the human operator to accept, reject or disregard the AI’s perspective

In this model, the reviewer remains fully responsible for the final decision. The AI system acts as a form of analytical support that helps surface signals that may not be easily detectable through visual inspection alone.

This approach can help strengthen review consistency while preserving human accountability.

Human Decision Support: AI as a Perspective, Not a Replacement

The most productive way to deploy AI in regulated environments is often as decision support rather than decision automation.

In practical terms, this means the system operates as a secondary analytical layer within the workflow.

How Human-First AI Review Works

1. Submission enters the workflow
   A user uploads an image, document or other form of evidence.

2. Authenticity analysis is performed
   Humanly analyses the submission for a range of authenticity signals that may indicate AI-generated or manipulated content.

3. Signals are surfaced to the reviewer
   The system presents findings that may warrant closer inspection.

4. Human reviewers remain in control
   The reviewer evaluates the AI’s perspective alongside their own judgement and operational policies.

5. Final decision stays with the human operator
   AI assists the process but does not determine the outcome.

This structure ensures organisations maintain clear human accountability, which remains critical for governance and regulatory alignment.

Why Perspective Matters in the Age of Synthetic Evidence

One useful way to think about AI assistance is as inviting a second perspective into the review process.

Human decision-makers already rely on multiple perspectives in many forms:

• Peer review
• Second-line oversight
• Specialist consultation
• Independent verification

AI analysis can function in a similar role — providing an additional viewpoint that helps reviewers consider whether a submission may require deeper scrutiny.

Importantly, the reviewer can choose whether to accept or disregard that perspective. The system exists to augment judgement, not override it.

This design principle helps maintain trust and transparency in environments where decisions carry regulatory or ethical implications.

Humanly’s Approach: Human First, AI Guarded

Humanly has been designed specifically for human-submitted evidence workflows, where organisations need to assess whether digital content is authentic.

Within this context, the platform is designed to help organisations:

• Analyse images or documents for signals associated with AI generation or manipulation
• Highlight submissions that may warrant closer review
• Support more consistent reviewer workflows
• Reduce the operational burden of manual authenticity checks

The objective is not to automate trust decisions. Instead, Humanly acts as an authenticity analysis layer that helps human reviewers navigate a more complex digital evidence landscape.

In practice, this means reviewers gain a form of analytical guardrail behind their decision-making, helping them approach submissions with greater situational awareness.

The Emerging Regulatory Question

As generative AI tools become more widely accessible, regulators and compliance teams are increasingly examining how organisations verify digital evidence and maintain decision integrity.

While regulatory expectations vary by sector, a consistent theme is emerging: organisations are expected to maintain robust controls over the information used in operational decisions.

In an environment where digital content can be synthetically generated, authenticity analysis may become a more important part of those controls.

This does not necessarily mean replacing human review with automation. Instead, it may involve equipping human reviewers with better analytical support so that decisions remain reliable and explainable.

From this perspective, the regulatory risk may not lie in using AI assistance responsibly. In some cases, the greater risk may come from relying solely on manual methods in a world where the underlying information environment has fundamentally changed.

Strengthening Operational Confidence in AI-Supported Workflows

Organisations evaluating AI adoption in regulated environments often benefit from focusing on three practical principles.

1. Maintain Human Decision Authority

Human reviewers should remain responsible for the final outcome. AI systems should provide analysis, not determinations.

2. Treat AI as Decision Support

AI outputs should be interpreted as signals or perspectives that inform human judgement rather than instructions to follow.

3. Build Transparent Review Processes

AI-assisted workflows should be designed so that reviewers understand what signals are being surfaced and how they contribute to the review process.

This model helps organisations balance innovation with governance, allowing AI to enhance operational capability without undermining accountability.

Bringing Authenticity Analysis Into Modern Operations

The information environment that organisations operate within is changing rapidly. As generative tools become more sophisticated, the boundary between naturally produced and synthetically generated content becomes harder to identify.

For teams responsible for reviewing evidence, this introduces new complexity.

Humanly’s role is to help organisations introduce authenticity intelligence into those workflows, enabling human reviewers to approach submissions with additional analytical support.

The result is a workflow that remains human-led but better equipped for a synthetic digital environment.

Explore Humanly

If your organisation reviews digital evidence as part of operational or regulatory workflows, Humanly can help introduce authenticity analysis into the process while keeping human reviewers firmly in control.

Enterprise teams:
Book a conversation to explore how Humanly can support your review workflows and help strengthen decision integrity.

Product-led users:
Create an account to see how Humanly analyses submissions for signals associated with AI-generated or manipulated content.

Healthcare systems have spent decades strengthening controls around data security, privacy and clinical governance. These efforts are essential. But a new category of risk is emerging that sits outside traditional frameworks.

That risk is evidence integrity.

Healthcare decision making increasingly relies on digital submissions. Images, referral letters, prescriptions, eligibility documents and supporting records are now central to claims processing, authorisation and access.

AI fundamentally alters the trustworthiness of these artefacts.

Beyond financial fraud

Healthcare fraud has traditionally been discussed in terms of cost. Inflated claims. Unnecessary procedures. Abuse of reimbursement systems.

AI introduces a more serious dimension.

When manipulated or synthetic evidence is used to obtain access to medication or treatment, the consequences extend to patient safety. Inappropriate access to prescription drugs, particularly high demand or controlled medications, creates risks that cannot be dismissed as administrative error.

Recent global demand for metabolic and weight loss drugs has highlighted this exposure. Where access decisions depend on digital documentation, the integrity of that evidence becomes critical.

Why this risk is difficult to see

Healthcare professionals are trained to assess clinical information, not the provenance of digital content. They are not forensic analysts. Nor should they be expected to be.

AI generated healthcare imagery and documents are often designed to appear plausible rather than perfect. They sit comfortably within expected ranges, making them difficult to challenge without specialised tools.

The result is a blind spot.

Digital access accelerates exposure

As healthcare systems expand digital access to improve efficiency and equity, reliance on remote evidence increases. This is positive, but it also magnifies the impact of misuse.

Manual review does not scale. Random audits are reactive. Blanket restrictions undermine access.

The only sustainable approach is layered assurance.

Authenticity as a safety control

Assessing whether healthcare related evidence appears genuine, edited or synthetic adds a new dimension to patient safety. It allows organisations to identify higher risk submissions without disrupting legitimate care.

This is not about denying access. It is about ensuring that access decisions are based on trustworthy information.

A future facing requirement

As AI continues to improve, healthcare systems that fail to address evidence integrity will face increasing pressure from regulators, auditors and public trust.

Evidence verification will become as fundamental as identity checks and data security.

Patient safety depends on it.

“In healthcare, manipulated evidence is not just fraud. It is a safety risk.”