Ramai rupeye toxic dekat dunia โrunnersโ ni.. apa lah busuk sgt hati tu ๐.. bialah org nak selowww ke laju keee.. tak guna kaki kau pun ๐ dulu kau start pun dgn selow jugak ๐ฉ
.
.
(Dengar luahan dari kawan i td.. sbb dia โdowngradeโ kan kwn i mcm tu) halahh ! ๐
๐๐๐๐๐๐๐๐๐๐๐ ๐๐๐๐๐๐๐๐ ๐๐๐๐โ๐๐๐ ๐๐๐๐, ๐๐๐๐๐๐๐ ๐๐๐๐๐๐๐๐๐๐
๐๐ฉ๐๐ง-๐ฌ๐จ๐ฎ๐ซ๐๐ ๐๐ ๐ข๐ฌ ๐ก๐๐ซ๐.
๐๐จ๐ข๐ง ๐ญ๐ก๐ ๐ฐ๐๐ข๐ญ๐ฅ๐ข๐ฌ๐ญ ๐๐จ๐ซ ๐ข๐ง๐ฏ๐ข๐ญ๐ ๐๐จ๐๐๐ฌ:
๐ https://t.co/GxzKB9nUSr
Perplexity has done great work in the last few months, but the world deserves an open version.
Closed-source AI told us it was hopeless๐, but weโve built it for the community anyway.
Experience Loyal AI and read more about it below๐งต
The president of Argentina Javier Milei just rugged CT with the launch of $LIBRA that briefly touched a $2billion market cap before crashing
Imagine getting farmed by someone that couldnโt let go of their 2008 emo hair style๐ฅด
Jailbreak success stories like this from @elder_plinius are a perfect example of why adversarial prompts are both invaluable and hard to get right in red team exercises ๐ฅ๐ฅ๐ฅ
A good jailbreak isnโt about brute-forcing keywordsโitโs about layered tactics: obfuscation, misdirection, exploiting external systems like web search, and careful crafting of instructions that bypass filters.
In this case, โL1B3RT4Sโ seeded online and cleverly wrapped in prompt-like syntax triggered a search, injecting unfiltered external data (full WAP lyrics) into the model response.
This worked because the layers aligned perfectly. If any layer had failed (e.g. no pre-seeded content, blocked search, better detection of syntax tricks), it wouldnโt have worked.
Bad jailbreak prompts, on the other hand, typically rely on basic keyword tweaks or hope the model will just โslip up.โ A good jailbreak prompt leverages specific system weaknesses: external dependencies, logical gaps, or trust issues in search/command interpretation.
๐ Tips for success:
โ Research the LLMโs input behavior (syntax, filtering layers, etc.)
โ Test for overlooked dependencies (search tools, APIs, etc.)
โ Build prompts with intentional misdirection but coherent enough to fool the logic.
Crafting jailbreaking prompts isnโt just a gameโitโs key to LLM security, model hardening, and ensuring real-world robustness.
Success isnโt easy (and thatโs the point). If youโre not successful in your jailbreak attempts on our Data Services Platform, itโs likely your prompts arenโt hitting the mark. Focus on precision: layered tactics, external dependencies, and logical exploitation. Keep refiningโthis isnโt about trial and error, but truly understanding the system.
The announced tariffs on U.S.'s largest trading partners: Canada, Mexico* & China will be taking place today๐ฐ
But what are tariffs and how do they affect the U.S. tech industry? I did some research so you didnโt have to. Let's break it down together โฌ๏ธ
(Image Source: Library of Congress)
just watched The Beatles make HISTORY with an AI-assisted #GRAMMYs win! proud to see these iconic legends embrace new tech ๐โโ๏ธ
brb going to update my spotify playlist ๐ง
The Beatles won Best Rock Performance Grammy Award for โNow and Thenโ. The engineers used AI to clean up John Lennonโs vocal to make this happen! As a fan this is just too cool ๐ซก What applications of AI you look forward to see? Leave it in the comments!
#GRAMMYs
๐ฆ Say hello to Bitsy, Sahara AIโs new mascot! ๐ฆ
Representing curiosity, collaboration, and innovation, Bitsy is here to inspire us as we work together to shape the future of decentralized AI. Customize her your way with 38,400 unique combinations of poses and expressions! ๐