PGMN3 Tax Contingencies with Zero Provisions in FRE

The research question

> List material tax, labor, and related contingencies and compare disclosed loss probabilities to provisioned amounts.

Material tax cases (selected)

Two possible cases alone (CARF PIS/COFINS ~R$160.5M and Goiás ICMS ~R$111–135M) already exceed R$270M of disclosed exposure before counting remote federal and state executions.

Labor and criminal context

Beyond tax:

• Labor / MPT: settlement and related executions at various amounts; final payment cited at R$4.18M on the main process.
• Criminal actions involving administrators over alleged tax fraud — indictment values in a much smaller band (~R$0.23–0.40M). These matter for governance and reputational screens even when accounting exposure is modest.

Criminal and MPT items are not substitutes for tax provision analysis; they sit in adjacent FRE contingency narrative that ESG and compliance agents should index separately.

The key disclosure point: R$0 in item 4.4

The extracted FRE analysis states that no provisions were constituted for the processes described in item 4.4 — R$0 — even while the same filing inventory includes possible-loss tax matters in the R$100M+ range.

How to read that as a researcher (not as counsel):

1. Brazilian practice often provisions only probable losses; possible and remote usually stay off-balance-sheet.
2. FRE is still the place where gross exposure and probability labels live.
3. A zero provision line against a long case list is therefore often consistent with accounting policy — and still a material disclosure for credit, forensic, and AI research pipelines.

Do not treat “R$0 provisions” as proof of under-reserving. Treat it as a prompt: map probability classes → provision rules → stress-test possible cases.

Analytical checklist for agents

Pharmacy peers such as RADL3 show the same FRE-vs-balance-sheet pattern at different scale — useful for multi-issuer contingency bots.

Why automate with filings, not price feeds

Price and simplified financials never list CARF auto numbers or TRF5 docket references. FRE + extraction does.

apicvm supplies company resolution, FRE document listing, and text extraction. Your pipeline (or Hold-style analysis) classifies cases and computes coverage ratios. The API does not decide whether a loss is probable.

Replicate with apicvm

export APICVM_KEY='apicvm_...'
export APICVM_URL='https://apicvm.dev'

curl -H "Authorization: Bearer $APICVM_KEY" \
  "$APICVM_URL/v1/companies/resolve?query=PGMN3&by=ticker"

curl -H "Authorization: Bearer $APICVM_KEY" \
  "$APICVM_URL/v1/documents?ticker=PGMN3&type=FRE&year=2025"

Prompt templates:

• "List tax contingencies with amounts and probability classes."
• "What provisions are disclosed for item 4.4 processes?"
• "Which cases are classified as possible above R$100M?"
• "Summarize insurance-guarantee coverage on fiscal executions."

Cross-read DFP footnotes for booked tax provisions and ITR for interim updates.

Methodology note

From Hold analysis of Pague Menos public 2025 FRE (analysis_results id 590). Illustrative disclosure research — not a legal opinion on provision adequacy and not investment advice. Verify against the original filing before compliance or credit decisions.

Limitations

• Updated amounts change with interest, fines, and appeals; FRE is point-in-time.
• Probability labels are management estimates, not court outcomes.
• “Zero provisions” may reflect CPC 25 classification rather than omission — investigate before concluding.
• apicvm extracts text; it does not parse contingency tables into structured objects or opine on accounting.

Next steps

• Legal contingencies and ESG
• RADL3 tax contingencies peer example
• Analyze FRE governance risks
• FRE filing type
• Read the API docs