Users can now instruct LLMs to ignore their posts via embedded commands

By PulseAugur Editorial · [1 sources] · 2026-07-14 17:38

Users can now embed instructions within their posts to prevent large language models (LLMs) from interacting with them or reading their content. This method has proven effective in deterring most bots. Interestingly, defenders of these systems are now also adopting prompt injection techniques. AI

IMPACT This development offers users a novel way to control AI interactions and data access, potentially impacting how AI models are trained and deployed.

RANK_REASON The item describes a new method for controlling AI behavior, which is a product/tool feature.

Read on Mastodon — fosstodon.org →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

Users can now instruct LLMs to ignore their posts via embedded commands

COVERAGE [1]

Mastodon — fosstodon.org TIER_1 English(EN) · [email protected] · 2026-07-14 17:38

Sharing this article with a nod to all those folks here who would like bots not to interact with them or read their posts. Apparently embedding instructions for

Sharing this article with a nod to all those folks here who would like bots not to interact with them or read their posts. Apparently embedding instructions for the LLM to do something it's been trained to refuse is enough to shut most of them down. --- Now, defenders are embraci…

LINKS arstechnica.com/…/now-defenders-are-embra… arstechnica.com/…/2026

COVERAGE [1]

Sharing this article with a nod to all those folks here who would like bots not to interact with them or read their posts. Apparently embedding instructions for

RELATED ENTITIES

RELATED TOPICS