← Insights home
The Horizon — Methodology

How this scan is produced

What sources we use, how we tier them, which lanes the scan covers, who reviews each issue, and what happens when we get something wrong. Plain language so you can decide what weight to put on the output.

01 — Lanes

The seven (plus three) areas we track

Every monthly scan aims to cover the lanes a Global 2000 GC is asked about. A quiet month in a lane is acceptable; a lane that goes empty issue after issue is a research gap and is treated as one.

AI
EU AI Act and member-state implementation, GPAI codes, US state AI laws, federal executive orders, court enforcement of AI rules.
Data & privacy
GDPR developments, UK DUAA, US state privacy laws, cross-border transfer mechanisms, China PIPL, India DPDP, sectoral data rules.
Sanctions & export controls
OFAC / OFSI / EU designations, dual-use and emerging-tech controls, secondary sanctions enforcement, deemed-export rules.
Competition & antitrust
DMA / DSA enforcement, FTC and DOJ actions, EU merger thresholds, CMA designations, competition reforms.
ESG & sustainability
CSRD, CS3D, SEC climate, ISSB adoption, supply-chain due diligence, modern slavery, green claims.
Employment & labour
EU Pay Transparency, Platform Work Directive, US state non-competes and AI-in-hiring rules, UK Employment Rights Bill.
Cyber & resilience
NIS2, DORA, CIRCIA, UK Cyber Security & Resilience Bill, sector cyber rules, incident notification regimes.
Tax
Pillar Two, DSTs, transfer-pricing developments, withholding rules that affect cross-border services.
Courts & enforcement
Material judgments, regulator fines, sanctions decisions, AI-litigation patterns, settlement benchmarks.
Other
Anything material that doesn't fit a standing lane (e.g. consumer law, IP, product safety).
02 — Regions

Coverage and inheritance

The page selector exposes All · UK · EU · Germany · France · US · APAC · ROW. The underlying tagging supports country-level codes too.

03 — Source-quality bar

Primary first, secondary as context

Every radar entry cites a source with a tier badge:

04 — Review

Human-in-the-loop, by design

Each issue is researched by a supervised agent against the rules above and submitted as a draft pull request. A qualified reviewer checks the sources, the date claims, and the framing before it is merged and published. The masthead on every issue records the named reviewer and the date the issue was last touched.

Issues are never auto-published. The product hook at the bottom of every issue makes the same point about per-company changes — the agent extends the reach of a lawyer's judgement, it does not replace it.

05 — Corrections

What happens when we get it wrong

Every correction we make to a published issue lands on the corrections page, dated and described. Where a deadline or scope moves materially, we also surface the change in the next issue's "Changes since last issue" callout so a returning reader sees the delta in context. If you spot an error, email hello@flank.ai with the issue title and the line.

06 — What we do not cover

The boundary, in plain terms

The Horizon is generalisable. It will not surface the change that touches a specific clause in your standard MSA, or the move that a competitor of yours quietly made in their terms. That gap is real and intentional — see section 05 of any issue for how a supervised agent closes it for a specific company. The scan is the right tool for the public signal. The agent is the right tool for the private translation.

Subscribe

The Horizon

Monthly. Delivered via Substack.

Subscribe on Substack
Flank

About Flank

Supervised AI agents for in-house legal teams.

flank.ai