Statistical watermark attribution from preserved public text

Establish whether the preserved text in the GROB corpus can support reliable statistical watermark attribution when no matched keyed detector is available.

Background

The report applies a text-provenance screen to the collected material and finds no known fixed watermark marker in the core wiki subset. Although literal provenance markers appear elsewhere in unrelated captured sources, they do not attribute the investigated traces. The report therefore leaves statistical watermark attribution unresolved because the preserved text could not be tested with a reliably applicable matched keyed detector.

References

Statistical watermark attribution remains unresolved because no matched keyed detector could be applied reliably to the preserved text.

— GROB: A Multi-Agent Architecture for Public-Trace Investigation of Candidate Agentic Activity  (2610.11467 - Bonfanti et al., 8 Oct 2026) in Section 5, subsection “Cross-family forensic checks”