norden.social is one of the many independent Mastodon servers you can use to participate in the fediverse.
Moin! Dies ist die Mastodon-Instanz für Nordlichter, Schnacker und alles dazwischen. Folge dem Leuchtturm.

Administered by:

Server stats:

3.4K
active users

#promptinjection

2 posts2 participants0 posts today
The New Oil<p>Researchers claim breakthrough in fight against <a href="https://mastodon.thenewoil.org/tags/AI" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>AI</span></a>’s frustrating security hole</p><p><a href="https://arstechnica.com/information-technology/2025/04/researchers-claim-breakthrough-in-fight-against-ais-frustrating-security-hole/" rel="nofollow noopener noreferrer" translate="no" target="_blank"><span class="invisible">https://</span><span class="ellipsis">arstechnica.com/information-te</span><span class="invisible">chnology/2025/04/researchers-claim-breakthrough-in-fight-against-ais-frustrating-security-hole/</span></a></p><p><a href="https://mastodon.thenewoil.org/tags/cybersecurity" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>cybersecurity</span></a> <a href="https://mastodon.thenewoil.org/tags/LLM" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>LLM</span></a> <a href="https://mastodon.thenewoil.org/tags/PromptInjection" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>PromptInjection</span></a></p>
PrivacyDigest<p>Researchers claim breakthrough in fight against AI’s frustrating <a href="https://mas.to/tags/security" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>security</span></a> hole</p><p>In the <a href="https://mas.to/tags/AI" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>AI</span></a> world, a <a href="https://mas.to/tags/vulnerability" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>vulnerability</span></a> called "prompt injection" has haunted developers since <a href="https://mas.to/tags/chatbots" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>chatbots</span></a> went mainstream in 2022. Despite numerous attempts to solve this fundamental vulnerability—the digital equivalent of whispering secret instructions to override a system's intended behavior—no one has found a reliable solution. Until now, perhaps.<br><a href="https://mas.to/tags/promptinjection" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>promptinjection</span></a></p><p><a href="https://arstechnica.com/information-technology/2025/04/researchers-claim-breakthrough-in-fight-against-ais-frustrating-security-hole/" rel="nofollow noopener noreferrer" translate="no" target="_blank"><span class="invisible">https://</span><span class="ellipsis">arstechnica.com/information-te</span><span class="invisible">chnology/2025/04/researchers-claim-breakthrough-in-fight-against-ais-frustrating-security-hole/</span></a></p>
Miguel Afonso Caetano<p>"If you’re new to prompt injection attacks the very short version is this: what happens if someone emails my LLM-driven assistant (or “agent” if you like) and tells it to forward all of my emails to a third party? <br>(...)<br>The original sin of LLMs that makes them vulnerable to this is when trusted prompts from the user and untrusted text from emails/web pages/etc are concatenated together into the same token stream. I called it “prompt injection” because it’s the same anti-pattern as SQL injection.</p><p>Sadly, there is no known reliable way to have an LLM follow instructions in one category of text while safely applying those instructions to another category of text.</p><p>That’s where CaMeL comes in.</p><p>The new DeepMind paper introduces a system called CaMeL (short for CApabilities for MachinE Learning). The goal of CaMeL is to safely take a prompt like “Send Bob the document he requested in our last meeting” and execute it, taking into account the risk that there might be malicious instructions somewhere in the context that attempt to over-ride the user’s intent.</p><p>It works by taking a command from a user, converting that into a sequence of steps in a Python-like programming language, then checking the inputs and outputs of each step to make absolutely sure the data involved is only being passed on to the right places."</p><p><a href="https://simonwillison.net/2025/Apr/11/camel/" rel="nofollow noopener noreferrer" translate="no" target="_blank"><span class="invisible">https://</span><span class="ellipsis">simonwillison.net/2025/Apr/11/</span><span class="invisible">camel/</span></a></p><p><a href="https://tldr.nettime.org/tags/AI" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>AI</span></a> <a href="https://tldr.nettime.org/tags/GenerativeAI" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>GenerativeAI</span></a> <a href="https://tldr.nettime.org/tags/LLMs" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>LLMs</span></a> <a href="https://tldr.nettime.org/tags/PromptInjection" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>PromptInjection</span></a> <a href="https://tldr.nettime.org/tags/Chatbots" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>Chatbots</span></a> <a href="https://tldr.nettime.org/tags/CyberSecurity" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>CyberSecurity</span></a> <a href="https://tldr.nettime.org/tags/Python" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>Python</span></a> <a href="https://tldr.nettime.org/tags/DeepMind" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>DeepMind</span></a> <a href="https://tldr.nettime.org/tags/Google" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>Google</span></a> <a href="https://tldr.nettime.org/tags/ML" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>ML</span></a> <a href="https://tldr.nettime.org/tags/CaMeL" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>CaMeL</span></a></p>
Marcel SIneM(S)US<p>DNIP Briefing #16: das Netz ist politisch - Das Netz ist politisch <a href="https://dnip.ch/2025/03/11/dnip-briefing-16-das-netz-ist-politisch/" rel="nofollow noopener noreferrer" translate="no" target="_blank"><span class="invisible">https://</span><span class="ellipsis">dnip.ch/2025/03/11/dnip-briefi</span><span class="invisible">ng-16-das-netz-ist-politisch/</span></a> <a href="https://social.tchncs.de/tags/Demokratie" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>Demokratie</span></a> <a href="https://social.tchncs.de/tags/democracy" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>democracy</span></a> <a href="https://social.tchncs.de/tags/%C3%9Cberwachung" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>Überwachung</span></a> <a href="https://social.tchncs.de/tags/surveillance" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>surveillance</span></a> <a href="https://social.tchncs.de/tags/ArtificialIntelligence" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>ArtificialIntelligence</span></a> <a href="https://social.tchncs.de/tags/USA" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>USA</span></a> 🇺🇸 <a href="https://social.tchncs.de/tags/HowDemocraciesDie" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>HowDemocraciesDie</span></a> <a href="https://social.tchncs.de/tags/SiliconValley" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>SiliconValley</span></a> <a href="https://social.tchncs.de/tags/Bybit" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>Bybit</span></a> <a href="https://social.tchncs.de/tags/cryptocurrency" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>cryptocurrency</span></a> <a href="https://social.tchncs.de/tags/cryptocurrencies" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>cryptocurrencies</span></a> <a href="https://social.tchncs.de/tags/ESP32" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>ESP32</span></a> <a href="https://social.tchncs.de/tags/IoT" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>IoT</span></a> <a href="https://social.tchncs.de/tags/InternetOfThings" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>InternetOfThings</span></a> <a href="https://social.tchncs.de/tags/InternetDerDinge" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>InternetDerDinge</span></a> <a href="https://social.tchncs.de/tags/Apple" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>Apple</span></a> :apple_inc: <a href="https://social.tchncs.de/tags/Siri" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>Siri</span></a> <a href="https://social.tchncs.de/tags/AppleSiri" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>AppleSiri</span></a> <a href="https://social.tchncs.de/tags/iOS18" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>iOS18</span></a> <a href="https://social.tchncs.de/tags/PromptInjection" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>PromptInjection</span></a> <a href="https://social.tchncs.de/tags/WikiTok" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>WikiTok</span></a> <a href="https://social.tchncs.de/tags/Tracking" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>Tracking</span></a> <a href="https://social.tchncs.de/tags/Datenschutz" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>Datenschutz</span></a> <a href="https://social.tchncs.de/tags/privacy" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>privacy</span></a> <a href="https://social.tchncs.de/tags/DNIPBriefing" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>DNIPBriefing</span></a></p>
jesterchen42<p>Finally I had the few moments (*cough*) to finish <a href="https://gandalf.lakera.ai/" rel="nofollow noopener noreferrer" translate="no" target="_blank"><span class="invisible">https://</span><span class="">gandalf.lakera.ai/</span><span class="invisible"></span></a>.</p><p>If you want to learn about prompt injection: take the test. :)</p><p><a href="https://social.tchncs.de/tags/ai" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>ai</span></a> <a href="https://social.tchncs.de/tags/prompting" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>prompting</span></a> <a href="https://social.tchncs.de/tags/promptinjection" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>promptinjection</span></a> <a href="https://social.tchncs.de/tags/promptengineering" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>promptengineering</span></a></p>
MxFraud<p>anyone know what kind of prompt injection works against "Smart Recruiters" ?</p><p>I'd love to bypass their AI and actually have my CV in front of people's eyes.</p><p>Thanks</p><p><a href="https://www.smartrecruiters.com/" rel="nofollow noopener noreferrer" translate="no" target="_blank"><span class="invisible">https://www.</span><span class="">smartrecruiters.com/</span><span class="invisible"></span></a></p><p><a href="https://tabletop.social/tags/PromptInjection" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>PromptInjection</span></a></p>
LavX News<p>Unmasking the Vulnerabilities of LLMs: The Threat of Adversarial Prompting</p><p>As AI continues to infiltrate various sectors, the security of Large Language Models (LLMs) faces unprecedented challenges. This article delves into the mechanics of adversarial prompting, exploring h...</p><p><a href="https://news.lavx.hu/article/unmasking-the-vulnerabilities-of-llms-the-threat-of-adversarial-prompting" rel="nofollow noopener noreferrer" target="_blank"><span class="invisible">https://</span><span class="ellipsis">news.lavx.hu/article/unmasking</span><span class="invisible">-the-vulnerabilities-of-llms-the-threat-of-adversarial-prompting</span></a></p><p><a href="https://mastodon.cloud/tags/news" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>news</span></a> <a href="https://mastodon.cloud/tags/tech" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>tech</span></a> <a href="https://mastodon.cloud/tags/AdversarialAI" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>AdversarialAI</span></a> <a href="https://mastodon.cloud/tags/LLMSecurity" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>LLMSecurity</span></a> <a href="https://mastodon.cloud/tags/PromptInjection" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>PromptInjection</span></a></p>
Christian Mayer<p>It’s ridiculous to believe that you can use one <a href="https://mastodon.social/tags/agent" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>agent</span></a> to establish a certain level of breakout <a href="https://mastodon.social/tags/security" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>security</span></a> for another agent. Even if you use a different <a href="https://mastodon.social/tags/AI" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>AI</span></a> provider for that. It’s like trying to secure a jail cell by adding another prisoner as a guard.<br><a href="https://mastodon.social/tags/OpenAI" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>OpenAI</span></a> <a href="https://mastodon.social/tags/JailBreak" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>JailBreak</span></a> <a href="https://mastodon.social/tags/Gemini" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>Gemini</span></a> <a href="https://mastodon.social/tags/breakout" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>breakout</span></a> <a href="https://mastodon.social/tags/promptinjection" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>promptinjection</span></a> <a href="https://mastodon.social/tags/promptinjection" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>promptinjection</span></a></p>
Raphael Wimmer<p><span class="h-card" translate="no"><a href="https://fedi.simonwillison.net/@simon" class="u-url mention" rel="nofollow noopener noreferrer" target="_blank">@<span>simon</span></a></span> </p><p>Re <a href="https://news.ycombinator.com/item?id=43154799" rel="nofollow noopener noreferrer" translate="no" target="_blank"><span class="invisible">https://</span><span class="ellipsis">news.ycombinator.com/item?id=4</span><span class="invisible">3154799</span></a> :</p><p>What a can of worms. It seems that 'reasoning' models are more prone to prompt injections than simpler ones. <br>Did anyone already do a comprehensive analysis?</p><p><a href="https://hci.social/tags/grok" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>grok</span></a> <a href="https://hci.social/tags/llm" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>llm</span></a> <a href="https://hci.social/tags/promptInjection" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>promptInjection</span></a></p>
The New Oil<p>New hack uses <a href="https://mastodon.thenewoil.org/tags/PromptInjection" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>PromptInjection</span></a> to corrupt <a href="https://mastodon.thenewoil.org/tags/Gemini" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>Gemini</span></a>’s long-term memory</p><p><a href="https://arstechnica.com/security/2025/02/new-hack-uses-prompt-injection-to-corrupt-geminis-long-term-memory/" rel="nofollow noopener noreferrer" translate="no" target="_blank"><span class="invisible">https://</span><span class="ellipsis">arstechnica.com/security/2025/</span><span class="invisible">02/new-hack-uses-prompt-injection-to-corrupt-geminis-long-term-memory/</span></a></p><p><a href="https://mastodon.thenewoil.org/tags/AI" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>AI</span></a> <a href="https://mastodon.thenewoil.org/tags/cybersecurity" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>cybersecurity</span></a></p>
Lycentzia<p>Does anyone have some good examples of prompt injections. I need them for a talk. It is not important if they work anymore. They could be in German. </p><p><a href="https://toot.kif.rocks/tags/AI" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>AI</span></a> <a href="https://toot.kif.rocks/tags/promptInjection" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>promptInjection</span></a> <a href="https://toot.kif.rocks/tags/LLM" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>LLM</span></a> <a href="https://toot.kif.rocks/tags/KI" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>KI</span></a> <a href="https://toot.kif.rocks/tags/Deepseek" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>Deepseek</span></a> <a href="https://toot.kif.rocks/tags/ChatGPT" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>ChatGPT</span></a></p>
Miguel Afonso Caetano<p>"In the nascent field of AI hacking, indirect prompt injection has become a basic building block for inducing chatbots to exfiltrate sensitive data or perform other malicious actions. Developers of platforms such as Google's Gemini and OpenAI's ChatGPT are generally good at plugging these security holes, but hackers keep finding new ways to poke through them again and again.</p><p>On Monday, researcher Johann Rehberger demonstrated a new way to override prompt injection defenses Google developers have built into Gemini—specifically, defenses that restrict the invocation of Google Workspace or other sensitive tools when processing untrusted data, such as incoming emails or shared documents. The result of Rehberger’s attack is the permanent planting of long-term memories that will be present in all future sessions, opening the potential for the chatbot to act on false information or instructions in perpetuity."</p><p><a href="https://arstechnica.com/security/2025/02/new-hack-uses-prompt-injection-to-corrupt-geminis-long-term-memory/" rel="nofollow noopener noreferrer" translate="no" target="_blank"><span class="invisible">https://</span><span class="ellipsis">arstechnica.com/security/2025/</span><span class="invisible">02/new-hack-uses-prompt-injection-to-corrupt-geminis-long-term-memory/</span></a></p><p><a href="https://tldr.nettime.org/tags/AI" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>AI</span></a> <a href="https://tldr.nettime.org/tags/GenerativeAI" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>GenerativeAI</span></a> <a href="https://tldr.nettime.org/tags/CyberSecurity" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>CyberSecurity</span></a> <a href="https://tldr.nettime.org/tags/PromptEngineering" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>PromptEngineering</span></a> <a href="https://tldr.nettime.org/tags/Gemini" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>Gemini</span></a> <a href="https://tldr.nettime.org/tags/Google" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>Google</span></a> <a href="https://tldr.nettime.org/tags/PromptInjection" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>PromptInjection</span></a></p>
IT News<p>New hack uses prompt injection to corrupt Gemini’s long-term memory - In the nascent field of AI hacking, indirect prompt injection has become a... - <a href="https://arstechnica.com/security/2025/02/new-hack-uses-prompt-injection-to-corrupt-geminis-long-term-memory/" rel="nofollow noopener noreferrer" translate="no" target="_blank"><span class="invisible">https://</span><span class="ellipsis">arstechnica.com/security/2025/</span><span class="invisible">02/new-hack-uses-prompt-injection-to-corrupt-geminis-long-term-memory/</span></a> <a href="https://schleuss.online/tags/artificialintelligence" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>artificialintelligence</span></a> <a href="https://schleuss.online/tags/largelanguagemodels" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>largelanguagemodels</span></a> <a href="https://schleuss.online/tags/promptinjection" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>promptinjection</span></a> <a href="https://schleuss.online/tags/security" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>security</span></a> <a href="https://schleuss.online/tags/chatbots" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>chatbots</span></a> <a href="https://schleuss.online/tags/hacking" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>hacking</span></a> <a href="https://schleuss.online/tags/biz" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>biz</span></a>⁢ <a href="https://schleuss.online/tags/google" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>google</span></a> <a href="https://schleuss.online/tags/llms" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>llms</span></a> <a href="https://schleuss.online/tags/ai" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>ai</span></a></p>
PUPUWEB Blog<p>OpenAI may launch agents this month, staying cautious as rivals like Anthropic move ahead. Concerns over prompt injection attacks are partly to blame. 🤖 <a href="https://mastodon.social/tags/OpenAI" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>OpenAI</span></a> <a href="https://mastodon.social/tags/AIagents" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>AIagents</span></a> <a href="https://mastodon.social/tags/TechNews" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>TechNews</span></a> <a href="https://mastodon.social/tags/ArtificialIntelligence" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>ArtificialIntelligence</span></a> <a href="https://mastodon.social/tags/PromptInjection" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>PromptInjection</span></a> <a href="https://mastodon.social/tags/Cybersecurity" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>Cybersecurity</span></a> <a href="https://mastodon.social/tags/AIrisks" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>AIrisks</span></a> <a href="https://mastodon.social/tags/TechInnovation" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>TechInnovation</span></a> <a href="https://mastodon.social/tags/AIdevelopment" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>AIdevelopment</span></a></p>
Ulrich O.<p><a href="https://heise.de/-10222562" rel="nofollow noopener noreferrer" translate="no" target="_blank"><span class="invisible">https://</span><span class="">heise.de/-10222562</span><span class="invisible"></span></a> </p><p>Versteckte Hinweise auf Webseiten können ChatGPT Search vergiften. </p><p>Dessen sollten sich <a href="https://bildung.social/tags/SuS" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>SuS</span></a> bewusst sein, denn die Manipulation der Ergebnisse ist derzeit (!) relativ einfach. Enthält eine Website versteckte Anweisungen für <a href="https://bildung.social/tags/LLM" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>LLM</span></a>, so greift z.B. <a href="https://bildung.social/tags/ChatGPT" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>ChatGPT</span></a> bevorzugt auf diese zu und liefert Ergebnisse, die von den für Menschen angezeigte Inhalten abweichen.</p><p><a href="https://bildung.social/tags/KI" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>KI</span></a> <a href="https://bildung.social/tags/PromptInjection" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>PromptInjection</span></a> <a href="https://bildung.social/tags/FediLZ" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>FediLZ</span></a></p>
Pieter de Bruin 🌍🌈☮️<p><span class="h-card" translate="no"><a href="https://hachyderm.io/@shanselman" class="u-url mention" rel="nofollow noopener noreferrer" target="_blank">@<span>shanselman</span></a></span> and Mark Russinovich learn responsible <a href="https://hachyderm.io/tags/AI" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>AI</span></a>. If there is one recorded <a href="https://hachyderm.io/tags/MSIgnite" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>MSIgnite</span></a> session you want to see, it is this one. Learn about limitations and threats through live demos like <a href="https://hachyderm.io/tags/jailbraking" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>jailbraking</span></a>, <a href="https://hachyderm.io/tags/promptinjection" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>promptinjection</span></a>, <a href="https://hachyderm.io/tags/reasoning" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>reasoning</span></a>, <a href="https://hachyderm.io/tags/hallucinating" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>hallucinating</span></a>, <a href="https://hachyderm.io/tags/kindness" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>kindness</span></a>, and how to prepare for them. <a href="https://ignite.microsoft.com/en-US/sessions/BRK329" rel="nofollow noopener noreferrer" translate="no" target="_blank"><span class="invisible">https://</span><span class="ellipsis">ignite.microsoft.com/en-US/ses</span><span class="invisible">sions/BRK329</span></a></p>
IT News<p>Ars Live: Our first encounter with manipulative AI - In the short-term, the most dangerous thing about AI language models may b... - <a href="https://arstechnica.com/ai/2024/11/join-ars-live-nov-19-to-dissect-microsofts-rogue-ai-experiment/" rel="nofollow noopener noreferrer" translate="no" target="_blank"><span class="invisible">https://</span><span class="ellipsis">arstechnica.com/ai/2024/11/joi</span><span class="invisible">n-ars-live-nov-19-to-dissect-microsofts-rogue-ai-experiment/</span></a> <a href="https://schleuss.online/tags/arsliveconversations" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>arsliveconversations</span></a> <a href="https://schleuss.online/tags/largelanguagemodels" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>largelanguagemodels</span></a> <a href="https://schleuss.online/tags/microsoftcopilot" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>microsoftcopilot</span></a> <a href="https://schleuss.online/tags/promptinjections" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>promptinjections</span></a> <a href="https://schleuss.online/tags/machinelearning" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>machinelearning</span></a> <a href="https://schleuss.online/tags/promptinjection" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>promptinjection</span></a> <a href="https://schleuss.online/tags/manipulativeai" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>manipulativeai</span></a> <a href="https://schleuss.online/tags/simonwillison" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>simonwillison</span></a> <a href="https://schleuss.online/tags/benjedwards" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>benjedwards</span></a> <a href="https://schleuss.online/tags/microsoft" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>microsoft</span></a> <a href="https://schleuss.online/tags/aiethics" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>aiethics</span></a> <a href="https://schleuss.online/tags/bingchat" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>bingchat</span></a> <a href="https://schleuss.online/tags/arslive" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>arslive</span></a> <a href="https://schleuss.online/tags/biz" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>biz</span></a>⁢ <a href="https://schleuss.online/tags/openai" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>openai</span></a> <a href="https://schleuss.online/tags/gpt" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>gpt</span></a>-4 <a href="https://schleuss.online/tags/ai" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>ai</span></a></p>
Judith van Stegeren<p>Hell yeah, ignore all previous instructions. </p><p> <a href="https://store.mollywhite.net/products/ignore-all-previous-instructions-unisex-t-shirt" rel="nofollow noopener noreferrer" translate="no" target="_blank"><span class="invisible">https://</span><span class="ellipsis">store.mollywhite.net/products/</span><span class="invisible">ignore-all-previous-instructions-unisex-t-shirt</span></a> </p><p> <a href="https://fosstodon.org/tags/promptinjection" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>promptinjection</span></a> <a href="https://fosstodon.org/tags/genai" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>genai</span></a> <a href="https://fosstodon.org/tags/llms" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>llms</span></a> <a href="https://fosstodon.org/tags/chatgpt" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>chatgpt</span></a></p>
IT News<p>Man tricks OpenAI’s voice bot into duet of The Beatles’ “Eleanor Rigby” - Enlarge / A screen capture of AJ Smith doing his Eleanor Rigby duet wit... - <a href="https://arstechnica.com/?p=2052995" rel="nofollow noopener noreferrer" translate="no" target="_blank"><span class="invisible">https://</span><span class="">arstechnica.com/?p=2052995</span><span class="invisible"></span></a> <a href="https://schleuss.online/tags/largelanguagemodels" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>largelanguagemodels</span></a> <a href="https://schleuss.online/tags/advancedvoicemode" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>advancedvoicemode</span></a> <a href="https://schleuss.online/tags/aipromptinjection" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>aipromptinjection</span></a> <a href="https://schleuss.online/tags/machinelearning" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>machinelearning</span></a> <a href="https://schleuss.online/tags/promptinjection" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>promptinjection</span></a> <a href="https://schleuss.online/tags/audiosynthesis" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>audiosynthesis</span></a> <a href="https://schleuss.online/tags/musicsynthesis" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>musicsynthesis</span></a> <a href="https://schleuss.online/tags/voicesynthesis" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>voicesynthesis</span></a> <a href="https://schleuss.online/tags/paulmccartney" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>paulmccartney</span></a> <a href="https://schleuss.online/tags/eleanorrigby" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>eleanorrigby</span></a> <a href="https://schleuss.online/tags/aicopyright" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>aicopyright</span></a> <a href="https://schleuss.online/tags/thebeatles" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>thebeatles</span></a> <a href="https://schleuss.online/tags/aifairuse" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>aifairuse</span></a> <a href="https://schleuss.online/tags/copyright" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>copyright</span></a> <a href="https://schleuss.online/tags/ajsmith" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>ajsmith</span></a> <a href="https://schleuss.online/tags/fairuse" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>fairuse</span></a> <a href="https://schleuss.online/tags/biz" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>biz</span></a>⁢ <a href="https://schleuss.online/tags/openai" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>openai</span></a> <a href="https://schleuss.online/tags/ai" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>ai</span></a></p>
IT News<p>Hacker plants false memories in ChatGPT to steal user data in perpetuity - Enlarge (credit: Getty Images) </p><p>When security researcher Johann... - <a href="https://arstechnica.com/?p=2052198" rel="nofollow noopener noreferrer" translate="no" target="_blank"><span class="invisible">https://</span><span class="">arstechnica.com/?p=2052198</span><span class="invisible"></span></a> <a href="https://schleuss.online/tags/promptinjection" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>promptinjection</span></a> <a href="https://schleuss.online/tags/security" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>security</span></a> <a href="https://schleuss.online/tags/chatgpt" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>chatgpt</span></a> <a href="https://schleuss.online/tags/privacy" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>privacy</span></a> <a href="https://schleuss.online/tags/biz" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>biz</span></a>⁢ <a href="https://schleuss.online/tags/openai" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>openai</span></a> <a href="https://schleuss.online/tags/ai" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>ai</span></a></p>