Vane detects what matters.
The difference is alpha.
Large language models are trained on the entire internet. They produce fluent summaries that sound authoritative but miss the specific, actionable signals that drive decisions. They can't distinguish between boilerplate risk language and genuine escalation.
| Vane | Off-the-shelf AI | |
|---|---|---|
| Source material | SEC filings, parsed & normalized | General internet corpus |
| Analysis type | Semantic shift detection | Summary generation |
| Output | Actionable escalation alerts | Generic descriptions |
| Temporal context | Before/after comparisons | Point-in-time snapshot |
| Industry benchmarks | Peer-normalized scoring | No context |
Escalation detected: Regulatory risk shifted from hypothetical to realized.
Q3 → Q4 change: "may face challenges" became "is facing regulatory action." Jurisdiction specified (EU). Timeline added (restructuring required by Q2).
Confidence: HighRecommended: Review MD&A section 4.2, Legal Proceedings.
The company faces regulatory challenges in certain jurisdictions that could impact operations. Management is monitoring the situation and will take appropriate actions as needed. Investors should consider regulatory risk as part of their investment decision.
No escalation detected. No temporal comparison. No specific action.
We maintain a corpus of each company's Risk Factors across every filing period. This lets us detect changes that point-in-time analysis misses entirely.
Our models identify specific escalation patterns: modal verb shifts, tense changes, specificity markers, new named entities, legal terminology.
Every score is normalized against industry peers. Know if a company is early, in-line, or lagging in disclosing risks others have already flagged.