For more than two decades, a quiet piece of software has shaped what a serious injury is worth in the United States. Its name is Colossus. Most claimants have never heard of it. Most claimants will never know it ran the numbers on their case. As generative AI moves into claim handling on top of that legacy infrastructure, the opacity is about to get much worse, and ASEAN regulators have a narrow window to choose a different path.
I practice personal injury law in Utah and Idaho, primarily on commercial trucking and serious-injury cases. The first thing my clients learn, often the hard way, is that the human adjuster they spoke to on the phone was not the entity deciding their case. Software was. AI is now sitting on top of that software. The result is a system that compresses payouts, accelerates low-ball offers, and is increasingly difficult to challenge through traditional discovery.
What “claim adjudication software” actually does
Colossus, originally developed in Australia and brought to the U.S. market in 1992, has been deployed at major U.S. property-and-casualty carriers since the mid-1990s. It is now owned by DXC Technology, the company formed from the 2017 merger of Computer Sciences Corporation and HPE Enterprise Services. Its peer product, Claims Outcome Advisor (now owned by Verisk and also marketed as Liability Navigator), followed a similar deployment curve. They take a structured input (injuries, treatment codes, demographics, and jurisdiction) and produce a payout range.
Internal documents released through litigation in the 2000s, most prominently in the Allstate “McKinsey” record and former-employee testimony from Mark Romano, established that Colossus had been tuned to produce settlement evaluations measurably lower than pre-deployment baselines, and that adjusters were measured against the tuned output. Allstate has consistently denied the broader characterization. The system and its successors remain in active deployment across the U.S. carrier market.
That part is decades old. What is new is the AI layer being grafted on top of it. Three additions, in particular, have changed the practitioner experience in the last 24 months:
- Large language model summarization of medical records, often surfaced to adjusters in place of the underlying records. Errors and omissions in those summaries flow downstream into the valuation.
- Automated triage of incoming claims, which routes certain categories of injuries (soft tissue, certain neurological presentations, and chronic-pain claims) into rapid low-offer pipelines before any human has reviewed the file.
- Pattern-matching tools that flag claimants for “fraud likelihood” based on signals (provider used, treatment cadence, and geography) that correlate with legitimate patient choices, not actual fraud.
The algorithm itself is not the policy problem. The opacity around it is.
What practitioners are seeing
In day-to-day litigation, three patterns now show up consistently.
Medical-record summaries that omit dispositive findings
A neurosurgeon’s note describing a positive Spurling test, for example, may not appear in a one-page summary fed to the adjuster. The valuation that follows reflects a less serious injury than the records describe. Discovery is the only reliable way to surface this, which means the lower-value, lower-leverage claims (the ones least likely to litigate) absorb the discount silently.
Rapid low-offer pipelines that bypass adjuster discretion
Claimants in certain injury categories receive a settlement offer generated within hours of submission, before a human adjuster has substantively reviewed the file. The offer cites generic policy language and a low number. The actual decision was made by a model.
Parallel patterns are surfacing in U.S. health-insurance litigation. Pending federal class actions against UnitedHealth’s nH Predict model (Estate of Lokken v. UnitedHealth Group, D. Minn., filed 2023) and against Cigna’s PXDX automated review system (E.D. Cal. and D. Conn., filed 2023) allege coverage denials issued without meaningful human review. Health and disability coverage is not P&C bodily injury, but the design pattern (model-first, human-after, sometimes never) is the same.
Fraud flags that survive after they are debunked
Once a claim is flagged, the flag tends to follow the claimant into every adjacent system at the insurer. Removing it requires escalation through channels most claimants do not know exist.
These are not edge cases. They are the median experience in a meaningful slice of U.S. personal injury practice in 2026.
Why ASEAN faces a faster version of the same problem
The U.S. insurance market is bolting AI onto an analog regulatory frame designed around paper records and human adjusters. ASEAN insurance markets, by contrast, are digital-first. Singapore’s Monetary Authority issued its FEAT Principles in 2018 and follow-on AI risk management guidance in 2025. Malaysia’s Bank Negara opened Digital Insurer and Takaful Operator licensing in 2025 under the Financial Sector Blueprint. Indonesia’s OJK is rolling out digital insurance product rules and a national policies database. The Philippines Insurance Commission is co-issuing guidance with the National Privacy Commission on privacy-enhancing technologies in insurance. The infrastructure is being built and rebuilt now, on cloud-native platforms, with AI in scope from day one.
That sequencing is, paradoxically, an advantage if regulators move quickly. The U.S. is litigating its way to AI insurance accountability state by state, more than two decades late. ASEAN can write the rules before deployment scales.
Three regulatory levers that would matter
Three concrete moves would address the worst pathologies of the U.S. experience, drawn from what I see in practice.
1. Mandatory disclosure of algorithmic involvement
Any claim denial, valuation, or fraud flag in which a model contributed to the outcome should be disclosed to the claimant in writing, naming the model, the version, and the input categories used. This is not a “right to explanation” in the abstract European sense. It is a procedural floor that allows a claimant or their counsel to ask the right discovery questions.
2. A documented human-review threshold
Above a defined claim value or severity threshold, model output should be advisory only. A licensed adjuster should sign the file, and the sign-off should attest that the adjuster reviewed the underlying records, not the model’s summary. This sounds obvious. In current U.S. practice, it is the exception.
3. Auditability of training and weighting data
Regulators should have standing to audit the training data, weighting, and update cadence of any model deployed in claim handling. The dominant U.S. analog (state insurance commissioners receiving rate filings) was not designed for algorithmic claim adjudication. Colorado is the leading exception. Its SB21-169 and 2025 implementing regulation extend algorithmic governance and bias-testing duties to claim handling. The NAIC Model Bulletin on AI, adopted by roughly half of U.S. states, extends documentation and validation duties across the insurance lifecycle. Most states still lack substantive audit authority over claim software. ASEAN regulators can build the audit pipeline into licensure now, while the deployment surface is still small.
The deeper point
This is not really an AI story. It is a digital public infrastructure story. Insurance is, in every developed market, a quasi-public good. It is regulated, subsidized at the edges, and central to economic resilience. When the adjudication layer of that public good is privatized into opaque software, the social contract underneath insurance erodes regardless of whether any specific outcome is “wrong.”
The U.S. has spent two decades reverse-engineering Colossus through litigation. ASEAN does not have to repeat that exercise. The infrastructure is being chosen now, the procurement contracts are being signed now, and the regulatory frame is, in most markets, still movable. The window will not stay open for long.
Kigan Martineau is a partner at BAM Personal Injury Lawyers (Benzion & Martineau Injury Law, PLLC), serving clients across Utah and Idaho from offices in St. George, Murray, and Meridian. His practice focuses on complex motor vehicle collisions, commercial trucking cases, and serious personal injury.
TNGlobal INSIDER publishes contributions relevant to entrepreneurship and innovation. You may submit your own original or published contributions subject to editorial discretion.
Featured image: Albert Stoynov on Unsplash
Unlocking the AI supercycle: Why infrastructure will determine Asia’s AI future