Poisoning the Knowledge Base - Revision history

Tom: /* Poisoning your Own Sites */

2026-01-18T05:50:04Z

‎Poisoning your Own Sites

Tom: /* Poisoning your Own Sites */

2026-01-18T05:47:32Z

‎Poisoning your Own Sites

Tom: /* Poisoning the Local Base */

2026-01-18T05:47:10Z

‎Poisoning the Local Base

Tom: /* Problems */

2025-12-20T01:17:42Z

‎Problems

Tom: /* =Poisoning the Global Base */

2025-12-19T21:37:58Z

‎=Poisoning the Global Base

Tom: /* Context */

2025-12-19T21:37:30Z

‎Context

Tom: /* Context */

2025-06-01T21:45:24Z

‎Context

Tom: /* Context */

2025-06-01T21:44:56Z

‎Context

Tom: Created page with "==Meme== Learning is just selecting your personal poison to inject into your own well of knowledge. ==Context== Learning should ideally be about expanding understanding, refi..."

2025-06-01T21:43:54Z

Created page with "==Meme== Learning is just selecting your personal poison to inject into your own well of knowledge. ==Context== Learning should ideally be about expanding understanding, refi..."

New page

==Meme==
Learning is just selecting your personal poison to inject into your own well of knowledge.

==Context==
Learning should ideally be about expanding understanding, refining insights, and improving accuracy—but in environments where data manipulation, misinformation, or AI bias exist, it can also be framed as a corrupting force if unchecked.

Your LinkedIn post touches on this idea of “poisoning the well”—where uncontrolled ingestion of personal data by AI models creates a skewed or exploitative system. If knowledge bases are built on unsanctioned data extraction, then learning might not be a pursuit of truth, but rather an assimilation of noise.

Are you thinking about intentional obfuscation techniques as a countermeasure? That’s an area where digital sovereignty might take an aggressive turn
==References==

[[Category: Privacy]]
[[Category: Knowledge]]

@@ Line 13: / Line 13: @@
 For sites that do not want their data to be scooped up by the large AI companies there is an effort started to poison any reads for those AI bots.
-Their initiative, dubbed [https://rnsaffn.com/poison3/ Poison Fountain], asks website operators to add links to their websites that feed AI crawlers poisoned training data. It's been up and running in early January 2026,
+Their initiative, dubbed [https://rnsaffn.com/poison3/ Poison Fountain], asks website operators to add links to their websites that feed AI crawlers poisoned training data. It's been up and running in early January 2026/<ref>Thomas Claburn, ''AI industry insiders launch site to poison the data that feeds them'' 2026-01-11 https://www.theregister.com/2026/01/11/industry_insiders_seek_to_poison/</ref>
 ===Poisoning the Global Base===

@@ Line 10: / Line 10: @@
 Thinking about intentional obfuscation techniques as a countermeasure? That’s an area where digital sovereignty might take an aggressive turn
-==Poisoning your Own Sites==
+===Poisoning your Own Sites===
 For sites that do not want their data to be scooped up by the large AI companies there is an effort started to poison any reads for those AI bots.
@@ Line 22: / Line 22: @@
 Israel, China and other countries are determined to change the bias in their own favor.
 * [https://www.misbar.com/en/editorial/2025/10/04/israel-contracts-new-firm-train-chatgpt-push-pro-israel-narratives-toward-gen-z-audiences Israel Contracts New Firm To Train ChatGPT To Push Pro-Israel Narratives Toward Gen Z Audiences] 2025-10-04
 ==Solutions==
 The web will shrink back into a collection of documents that are used to inform the [[Agent Internet]]. Some means of marking the credibility or provenance of this documents must evolve.

@@ Line 10: / Line 10: @@
 Thinking about intentional obfuscation techniques as a countermeasure? That’s an area where digital sovereignty might take an aggressive turn
 ===Poisoning the Global Base===
 The training data for the overall knowledge base can be caused by simple bias in the training data, or by deliberate attempts to alter the base.

@@ Line 17: / Line 17: @@
 Israel, China and other countries are determined to change the bias in their own favor.
 * [https://www.misbar.com/en/editorial/2025/10/04/israel-contracts-new-firm-train-chatgpt-push-pro-israel-narratives-toward-gen-z-audiences Israel Contracts New Firm To Train ChatGPT To Push Pro-Israel Narratives Toward Gen Z Audiences] 2025-10-04
 ==References==

@@ Line 10: / Line 10: @@
 Thinking about intentional obfuscation techniques as a countermeasure? That’s an area where digital sovereignty might take an aggressive turn
-===Poisoning the Global Base==
+===Poisoning the Global Base===
 The training data for the overall knowledge base can be caused by simple bias in the training data, or by deliberate attempts to alter the base.

@@ Line 4: / Line 4: @@
 ==Context==
 Learning should ideally be about expanding understanding, refining insights, and improving accuracy—but in environments where data manipulation, misinformation, or AI bias exist, it can also be framed as a corrupting force if unchecked.
+==Problems==
 The idea of “poisoning the well” is where uncontrolled ingestion of personal data by AI models creates a skewed or exploitative system. If knowledge bases are built on unsanctioned data extraction, then learning might not be a pursuit of truth, but rather an assimilation of noise.
 Thinking about intentional obfuscation techniques as a countermeasure? That’s an area where digital sovereignty might take an aggressive turn
 ==References==

@@ Line 5: / Line 5: @@
 Learning should ideally be about expanding understanding, refining insights, and improving accuracy—but in environments where data manipulation, misinformation, or AI bias exist, it can also be framed as a corrupting force if unchecked.
-Your LinkedIn post touches on this idea of “poisoning the well”—where uncontrolled ingestion of personal data by AI models creates a skewed or exploitative system. If knowledge bases are built on unsanctioned data extraction, then learning might not be a pursuit of truth, but rather an assimilation of noise.
+The idea of “poisoning the well” is —where uncontrolled ingestion of personal data by AI models creates a skewed or exploitative system. If knowledge bases are built on unsanctioned data extraction, then learning might not be a pursuit of truth, but rather an assimilation of noise.
 Are you thinking about intentional obfuscation techniques as a countermeasure? That’s an area where digital sovereignty might take an aggressive turn
 ==References==
 [[Category: Privacy]]
 [[Category: Knowledge]]