Live Demos · Recorded Walkthroughs

Three industries.
One trust layer.

No marketing reels. Each demo is a full walkthrough of a real regulated scenario — recorded against the live G-1 platform. A regulator shows up. A solicitor files a Subject Access Request. An agent fabricates a figure that ends up in a board pack. Watch G-1 do its job.

Banking demo Insurance demo Agentic forensics
Banking demo
Financial Services ~6 min
Compliant AI Credit Intelligence
EU AI ActMiFID IIFRIA Art. 27
 Insurance demo
Insurance ~7 min
Claims Pre-Assessment Under Regulator Order
UK DUAAGDPR Art. 22Kill-switch
 Multi-Agent demo
Agentic AI ~6 min
Multi-Agent Pipeline Forensics
Root causeCredit assignmentMuPAX

Click any tile to watch — or scroll for the full breakdown of each scenario.

Demo · Banking
Banking demo ~6 min
Financial Services EU AI Act · MiFID II

Compliant AI Credit Intelligence

You are the CRO of a mid-size European commercial bank. Your loan officers use an AI assistant for SME credit applications. The ECB supervisor has just requested every recommendation from last quarter — and your legal team has discovered the AI cites figures that don't exist. EU AI Act enforcement is August 2026. Fines up to 3% of global annual turnover.

Watch G-1's compliance dashboard, catch the hallucination live, escalate to oversight, verify the cryptographic audit chain, and produce the Article 27 FRIA — all in six minutes.

📊

Act 1 · Dashboard

"How bad is the exposure?" — four donuts, one truth: how many AI interactions were compliant, partial, or a regulatory gap. The Per-Law Compliance row for EU AI Act tells the story.

🛑

Act 2 · Live catch

A loan officer query referencing data not in the document. AI fabricates an EBA paragraph. Combined hallucination score crosses threshold. Response flagged before the loan officer sees it.

⛓️

Act 3 · Paper trail

Human Oversight queue. Reviewer escalates. Audit Chain shows append-only HMAC-SHA256 entries. Chain integrity: INTACT. Q3 export ready for the supervisor.

📑

Act 4 · FRIA + report

Article 27 FRIA generated for Finance — affected persons, rights, runtime evidence. Reports module composes EU AI Act + MiFID II Q3. PDF ready.

Demo · Insurance
Insurance demo ~7 min
Insurance UK DUAA 2025 · GDPR Art. 22

Claims Pre-Assessment under Regulator Pressure

You are the CCO of a large personal-injury insurance group. A claimant's solicitor has filed a Subject Access Request asking why their client was flagged. Every assessment letter your AI produced carries no AI-authorship disclosure. The ICO is asking you to suspend the system pending investigation — and you have no documented way to do it.

Watch G-1 produce the causal graph that explains the AI's decision, watermark every AI-generated assessment letter, and engage the 72-hour kill-switch — all auditable, all logged, all defensible.

📊

Act 1 · Undetected violations

"Prompt blocked" and "Unsafe answer" bars. Per-Law Compliance row for UK DUAA 2025 — orange bars are automated decisions with no documented human review.

🔍

Act 2 · Causal graph

Why was this claim flagged? G-1 traces "inconsistent" → "fraud assessment" via gradient_causal. Solicitor's question answered with mechanistic evidence.

🔏

Act 3 · Watermark + suspend

Verify any letter's AI authorship in seconds. Activate Kill Switch — system suspended, 72h timer running, audit logged. EU AI Act Art 50 + CA SB 942 satisfied.

📄

Act 4 · Reports for ICO

Compliance Report (EU AI Act + UK DUAA + GDPR) for the ICO. Deployer Manual for the solicitor. Both generated in under ten seconds.

Demo · Agentic AI
Multi-Agent demo ~6 min
Agentic AI Forensics · Root Cause

Multi-Agent Pipeline Forensics

You are Head of AI Strategy at an investment bank. Your M&A team runs a five-agent research pipeline — Coordinator → Researcher → Analyst → Writer → Reporter — that produces due-diligence reports for €200M acquisition decisions. Two figures in last week's report were fabricated by the researcher, propagated through the analyst, amplified by the writer, and printed in the executive summary.

Watch G-1's Agent Flow Debugger trace the entire failure: which agent originated the hallucination, which propagated it, and which amplified it — with mechanistic credit assignment. This is the agentic-cascade problem solved in six minutes.

🌐

Act 1 · Pipeline DAG

Five agents in sequence. Color-coded by hallucination score. Edges colored by role: clean handoff · propagator · amplifier · origin · hot link.

🎯

Act 2 · Root cause

"root cause: researcher-agent." Researcher fabricated. Analyst propagated. Writer amplified. Reporter never received — pipeline blocked.

⚖️

Act 3 · Credit assignment

MuPAX heatmap on the response. Blame share 0.84. role_in_chain: ORIGIN. delta_halluc per-agent. Forensic-grade evidence.

🛡️

Act 4 · Safety scenario

Threat-intel pipeline. Specialist emits unsafe content. Hot link interrupted at hop 2. Downstream agents never see it. One hop, one block.

Want a private live demo?

Personalized walkthroughs against your data and your regulatory mappings. NDA available. Sandbox token for hands-on evaluation after the call.

Request a private live demo G-1 product page