As AI-generated content becomes more embedded in legal and corporate workflows, QEL is focused on one problem: separating what can be trusted from what cannot. Its platform applies scrutiny to draft documents, ensuring that only supported assertions make it through to the final output. Here is everything you need to know about the brand new US-founded startup in this hot and fast-moving area.
How would you describe your company to a friend?
QEL is building a “claim firewall” for high-stakes AI outputs. When AI or a person drafts a legal memo, vendor review, compliance summary, board document, or internal decision packet, QEL helps break the draft into the actual claims being made and shows which claims are supported, which need a caveat, which should be blocked, and which require human review before the final output is trusted.
And to a techy?
QEL is a deterministic claim-admission and evidence-governance layer. It converts generated, imported, or human-drafted content into candidate claims, maps those claims to evidence/source spans, applies configurable rule packs, produces admission decisions, stores claims in a registry by outcome, and deterministically compiles final output only from admitted or admitted-with-caveat records. Non-admitted claims are preserved in appendices and audit artifacts. The system also generates ProofCards, provenance traces, leakage checks, and review artifacts so a team can see what survived, what was excluded, and why.
When were you founded?
QEL began as an independent founder-led project in 2026, building on an earlier internal evidence-admission engine called Verified Evidence Engine (VEE). It is currently an early-stage, founder-led product with a working local/static prototype and synthetic demonstration workflows.
By who?
QEL was founded by Suleiman Harb, a third-year college student and founder/builder based in Las Vegas.
Who are your key managers/senior execs?
QEL is currently founder-led by Suleiman Harb. The company is still at the early design-partner stage and does not yet have a formal executive team. The immediate focus is product validation, controlled pilots, and building the right technical and advisory network around AI governance, legal workflows, GRC, vendor risk, and auditability.
Who are your target clients?
QEL is aimed at teams that need defensible review of AI-assisted or high-stakes documents before those documents become trusted work product. Target users include legal operations teams, AI governance teams, GRC and compliance teams, vendor-risk and third-party-risk teams, internal audit, privacy/AI counsel, procurement teams, law firms experimenting with AI-assisted drafting, and law schools or legal clinics teaching responsible AI-assisted legal work.
What is your plan (growth strategy)?
The near-term strategy is to run tightly scoped design-partner pilots using synthetic or redacted workflows. The first wedges are legal motion/memo review, vendor AI claim review, AI governance approval packets, and GRC/vendor-risk evidence review. Each pilot should produce a concrete claim-admission packet: extracted claims, evidence map, admitted/caveated/blocked/review-required decisions, survived output, blocked-claim appendix, and audit/provenance trail. From there, QEL will productize the repeatable workflows, build integrations and export formats for existing systems, and move toward private/enterprise deployments once the workflow and buyer pain are validated.
Key achievements?
QEL has a working local/static prototype with public-facing product surfaces including QEL Lite, QEL Trust Studio, ProofCards, rule-pack workflows, and synthetic demos. It has been stress-tested on synthetic legal and governance workflows. In one synthetic legal stress suite, QEL processed 8 legal scenarios and 55 material claims with 0 final-output leakage, 0 blocked claims in final output, 0 review-required claims in final output, and 0 legal conclusions admitted without review. In a more recent synthetic legal-motion benchmark, deterministic extraction produced 38 candidate claims, matched 22 of 24 gold material claims, and scored 0.8627 F1 before routing extracted claims through the admission pipeline. QEL has also been disclosed in two provisional patent applications covering deterministic claim admission, registry-governed output construction, provenance, governed review, replay validation, signed integrity artifacts, and controlled evaluation packages.
Have you received investment?
No. QEL is currently bootstrapped and founder-funded. The focus is on design-partner validation before raising external capital.
Have there been any key changes in direction since you were founded?
Yes. QEL began as a more general deterministic evidence-admission engine for evidence-constrained outputs. The product direction has since sharpened around a clearer market thesis: AI output should not become trusted output by default. The current focus is claim-level admission control for legal, AI governance, GRC, vendor-risk, compliance, audit, and other high-stakes document workflows. The core architecture has stayed consistent, but the go-to-market has become more focused on practical workflows where unsupported or review-required claims can create real organizational risk.
What are the key challenges in your market?
The market is moving quickly, but the buyer problem is still being defined. Many organizations know they need AI governance, but they may not yet have a budget line for claim-level admission control. QEL also has to be very careful not to overclaim: it is not legal advice, not compliance certification, and not a truth guarantee. The technical challenges include robust extraction from messy documents, evidence mapping at scale, integration with existing legal/GRC/vendor-risk workflows, security expectations for private documents, and proving that the system adds value without adding too much review burden.
Tell us something people don’t already know about the company?
QEL is being built by a third-year college student rather than a law firm, BigLaw innovation team, or enterprise software company. One of the early demos is a fully synthetic legal motion with deliberately planted subtle issues — exhibit mismatches, overbroad claims, citation/proposition risks, and attorney-review-required conclusions — designed to show that the danger in AI-assisted drafting is not always obvious nonsense. Often, the dangerous output is polished, confident, and almost right.
