Announcement
Second Thought reads your draft locally, points out risky phrasing, and offers safer wording.
What it does not do, first. It advises and never stops a send. It is bypassable, by design and in one click. It runs on one surface, it works in English only, and its eight rules are a demonstration of the idea rather than a compliance programme.
The checker runs inside your browser tab. The first time you visit, it downloads once and stays in your browser cache. After that, every check happens on your own device with nothing sent anywhere.
A scan runs in stages, cheapest first. A context gate decides which rules apply to this recipient. A deterministic stage runs patterns and literal terms over the raw draft; it needs no external service, so the page is already working in its first second while the rest downloads.
The meaning-based stage runs only when a cue phrase is present, no negation governs that cue, and the sentence clears that rule’s similarity bar against its examples. All three conditions are needed. The checker encodes topic and vocabulary far more strongly than stance, so we cannot guarantee your data never leaves the US scores 0.978 against the data-residency examples while the actual promise scores 0.964. Similarity alone would put a card under the most careful sentence in the draft.
Section 7 of the NLRA protects most private-sector employees discussing pay and working conditions. Regulators have at times pressed for heightened scrutiny of monitoring technology on that basis; the specific NLRB General Counsel guidance from 2022 was rescinded in February 2025, and enforcement posture here moves with each administration, which is exactly why we did not want the architecture to depend on it.
In Germany, §87(1)(6) BetrVG gives a works council, where one exists, co-determination over technical devices for monitoring behaviour or performance, and German case law reads capability rather than intent as the trigger. Under EU and UK data-protection law, monitoring employee messages is the kind of processing supervisory authorities treat as likely high-risk, so a DPIA is normally required, and employee consent is a weak basis because of the power imbalance.
So: nothing you type leaves your computer, there is no endpoint that could receive it, there is no console showing anyone what anyone typed, there are no per-person scores, and the rule set excludes pay, hours, working conditions, and organizing by construction.
The site is static files, its content security policy permits connections to this origin only, and the network tab is the entire audit.
Eight rules ship in the demo. Six apply only to external recipients. Tone and language apply everywhere.
| Rule | Severity | Applies in |
|---|---|---|
| An unconditional promise about where data is storedresidency-promise | High | External only |
| An absolute claim about securityabsolute-security-claim | High | External only |
| A performance number stated as a guaranteeperformance-guarantee | Medium | External only |
| A delivery date offered as a commitmentdelivery-date-commitment | Medium | External only |
| A price or discount offered without approvalpricing-commitment | Medium | External only |
| Moving the conversation somewhere it is not recordedmoving-off-record | High | External only |
| This reads as criticism of a person rather than the workharsh-criticism | Medium | Internal and external |
| Strong languagestrong-language | Low | Internal and external |
Three items are in development. A browser extension to bring the same checks into live drafting. Rule authoring with import and export, so teams can tune and share their own rule sets. Rewriting in the writer’s own voice rather than fixed replacement text.
See the roadmap page for current status and technical detail.
Second Thought is a drafting aid. It does not provide legal or compliance advice and does not ensure compliance with any law, regulation, or company policy.