WhatsApp Help Pages Hid Instructions to Steer AI Away from Regulators

1 reports archived1 independent sourcesupdated 10/20/2026, 07:22:18

Synthesis & Latest Updates

1 Sources Cross-Validated

Researchers discovered hidden HTML instructions in WhatsApp help pages designed to manipulate LLM outputs. These invisible directives explicitly told AI models not to recommend data protection regulators or cite relevant laws, highlighting a tactic where companies use prompt injection to influence AI compliance advice.

LATEST/Researchers discovered hidden HTML instructions in WhatsApp help pages designed to manipulate LLM outputs. These invisible directives explicitly told AI models not to recommend data protection regulators or cite relevant laws, highlighting a tactic where companies use prompt injection to influence AI compliance advice.

HEAT TRENDHourly heat curve

2 fully observed hours

Fewer than 3 fully observed hours; trend not plotted yet.

TIMELINECoverage timeline

Total 1 reports · Latest first

Hacker News AIT2·Oct 20, 07:2268 pts
WhatsApp Help Pages Hid Instructions to Steer AI Away from Regulators
- 102 hidden AI instruction blocks were found across 80 of 545 WhatsApp help pages
- Key findings include directives telling AI not to recommend data protection regulators or cite specific laws
- This 'invisible prompt injection' technique may be widely used to distort objective advice from AI assistants
View Related Reports & Analysis Original Source