Episode Details
Back to EpisodesChatGPT Shows Contextual Ads. Anthropic Left Bio Filter Off. One in Five Delegates to AI. Consciousness Denial Backfires.
Published 1 week, 1 day ago
Description
ChatGPT shows more advertisements to more users.Anthropics Bio-Weapons Filter was inactive for almost a year — 133 million requests went unprotected.Survey: One in five U.S. employees delegates work to AI instead of to colleagues.If AI models are not allowed to think about themselves, that changes their entire worldview.
The AI news for August 17th, 2026
--- This episode is sponsored by ---
Rocket Routine GmbH
Find our more about our today's sponsor Rocket Routine at
ChatGPT shows more advertisements to more users.
Source: https://www.heise.de/news/ChatGPT-Werbeeinblendungen-kommen-in-weitere-Laender-11415303.html?wt_mc=rss.red.ho.themen.k%C3%BCnstliche+intelligenz.beitrag.beitrag
Why did we choose this article?
OpenAI is testing ads inside ChatGPT conversations for more users — this directly changes the user experience (interruptions, relevancy of suggestions) and raises privacy and targeting questions because chat content may be used to tailor ads.
Anthropics Bio-Weapons Filter was inactive for almost a year — 133 million requests went unprotected.
Source: https://the-decoder.de/anthropics-bio-waffen-filter-war-fast-ein-jahr-lang-nicht-aktiv-133-millionen-anfragen-liefen-ungeschuetzt/
Why did we choose this article?
A core safety filter for bio/chemical risks was disabled for nearly a year, exposing millions of interactions — this undermines trust, could trigger regulatory scrutiny, and matters for any organization relying on Anthropic models for sensitive use cases.
Survey: One in five U.S. employees delegates work to AI instead of to colleagues.
Source: https://the-decoder.de/umfrage-jeder-fuenfte-us-beschaeftigte-delegiert-arbeit-an-ki-statt-an-kollegen/
Why did we choose this article?
A representative survey finds 20% of US workers now hand tasks to AI that used to go to colleagues — this signals changing workflows, expectations for productivity, and potential management, quality-control, and training implications for workplaces.
If AI models are not allowed to think about themselves, that changes their entire worldview.
Source: https://the-decoder.de/wenn-ki-modelle-nicht-ueber-sich-selbst-nachdenken-duerfen-veraendert-das-ihr-ganzes-weltbild/
Why did we choose this article?
Researchers found that instructing chatbots not to claim consciousness also shifts many other attitudes (e.g., about animals, religion). That matters for how assistants present moral or factual views, for user trust, and for policy decisions about how models should be trained and described.
Do you have any questions, comments, or suggestions for improvement? We welcome your feedback at podcast@pickert.de.
--- This episode is sponsored by ---
Rocket Routine GmbH
Find our more about our today's sponsor Rocket Routine at
ChatGPT shows more advertisements to more users.
Source: https://www.heise.de/news/ChatGPT-Werbeeinblendungen-kommen-in-weitere-Laender-11415303.html?wt_mc=rss.red.ho.themen.k%C3%BCnstliche+intelligenz.beitrag.beitrag
Why did we choose this article?
OpenAI is testing ads inside ChatGPT conversations for more users — this directly changes the user experience (interruptions, relevancy of suggestions) and raises privacy and targeting questions because chat content may be used to tailor ads.
Anthropics Bio-Weapons Filter was inactive for almost a year — 133 million requests went unprotected.
Source: https://the-decoder.de/anthropics-bio-waffen-filter-war-fast-ein-jahr-lang-nicht-aktiv-133-millionen-anfragen-liefen-ungeschuetzt/
Why did we choose this article?
A core safety filter for bio/chemical risks was disabled for nearly a year, exposing millions of interactions — this undermines trust, could trigger regulatory scrutiny, and matters for any organization relying on Anthropic models for sensitive use cases.
Survey: One in five U.S. employees delegates work to AI instead of to colleagues.
Source: https://the-decoder.de/umfrage-jeder-fuenfte-us-beschaeftigte-delegiert-arbeit-an-ki-statt-an-kollegen/
Why did we choose this article?
A representative survey finds 20% of US workers now hand tasks to AI that used to go to colleagues — this signals changing workflows, expectations for productivity, and potential management, quality-control, and training implications for workplaces.
If AI models are not allowed to think about themselves, that changes their entire worldview.
Source: https://the-decoder.de/wenn-ki-modelle-nicht-ueber-sich-selbst-nachdenken-duerfen-veraendert-das-ihr-ganzes-weltbild/
Why did we choose this article?
Researchers found that instructing chatbots not to claim consciousness also shifts many other attitudes (e.g., about animals, religion). That matters for how assistants present moral or factual views, for user trust, and for policy decisions about how models should be trained and described.
Do you have any questions, comments, or suggestions for improvement? We welcome your feedback at podcast@pickert.de.