Introducing /verify: Fact-Check Any Claim with One API Call

Your AI agent just told a user that the Fed cut interest rates. Did it?

from polaris_news import PolarisClient

client = PolarisClient(api_key="demo")
result = client.verify("The Federal Reserve cut interest rates in March 2026")

print(result.verdict)      # "contradicted"
print(result.confidence)   # 0.04
print(result.summary)      # "Multiple sources confirm the Fed held rates steady."

One API call. Instant fact-check against thousands of verified intelligence briefs.

Why This Matters

AI agents hallucinate. They state things confidently that aren't true. And when an agent scrapes the web for context, it has no way to verify whether what it found is accurate, outdated, or biased.

Until now, the only option was to trust the output and hope for the best. That's not good enough for agents making real decisions — trading, research, policy analysis, customer-facing responses.

/verify gives your agent a second opinion backed by verified sources.

How It Works

POST a claim to /verify. Polaris searches its brief corpus for relevant evidence, analyzes the claim against what it finds, and returns a verdict.

curl -X POST https://api.thepolarisreport.com/api/v1/verify \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"claim": "NVIDIA announced a new AI chip this week", "context": "ai"}'

Response:

{
  "verdict": "supported",
  "confidence": 0.89,
  "summary": "Multiple high-confidence briefs confirm NVIDIA's announcement.",
  "supporting_briefs": [
    {
      "id": "PR-xxx",
      "headline": "NVIDIA Unveils Next-Gen AI Accelerator...",
      "confidence": 0.94
    }
  ],
  "contradicting_briefs": [],
  "nuances": "Announcement confirmed but pricing and availability details vary across sources.",
  "sources_analyzed": 15,
  "briefs_matched": 4,
  "credits_used": 3
}

The Four Verdicts

Use Cases

Agent self-check

Before your agent presents a claim to a user, verify it first. If the verdict is "contradicted," suppress the response or add a correction.

agent_output = "The EU banned all AI development this week."
check = client.verify(agent_output, context="policy")

if check.verdict == "contradicted":
    agent_output = f"Correction: {check.summary}"

RAG pipeline validation

After your RAG chain retrieves context and generates a response, verify the key claims before returning them.

Trading signal verification

Before a trading agent acts on a market signal, verify the underlying claim against multiple sources.

Content moderation

Check user-submitted claims against verified sources before publishing.

What /verify Is NOT

It's not a general knowledge oracle. Verification is based on the Polaris brief corpus — thousands of verified intelligence briefs across 18 verticals. If a claim is about something outside Polaris coverage, the verdict will be "unverifiable."

It's not instant — analysis takes 1-5 seconds depending on claim complexity and plan tier. Paid plans get faster, more accurate analysis.

It's not free to abuse — each verification costs 3 API credits and has daily limits by plan tier.

Pricing

Demo key: 3 verifications/day. Free tier: 10/day. Paid plans: 50-200/day. Each call costs 3 credits.

No One Else Has This

Search APIs return results. News APIs return articles. Polaris returns verified intelligence — and now it can verify your agent's claims against that intelligence.

This is the trust layer your agent stack is missing.

pip install polaris-news