Grok 2.0 takes the guardrails off AI image generation

By Andrew Tarantola Published August 14, 2024

Elon Musk’s xAI company has released two updated iterations of its Grok chatbot model, Grok-2 and Grok-2 mini. They promise improved performance over their predecessor, as well as new image-generation capabilities that will enable X (formerly Twitter) users to create AI imagery directly on the social media platform.

“We are excited to release an early preview of Grok-2, a significant step forward from our previous model, Grok-1.5, featuring frontier capabilities in chat, coding, and reasoning. At the same time, we are introducing Grok-2 mini, a small but capable sibling of Grok-2. An early version of Grok-2 has been tested on the LMSYS leaderboard under the name ‘sus-column-r,’” xAI wrote in a recent blog post. The new models are currently in beta and reserved for Premium and Premium+ subscribers, though the company plans to make them available through its Enterprise API later in the month.

Recommended Videos

The image-generation feature appears to be powered by the Flux.1 model developed by Black Forest Labs. While virtually every other image-generation system on the market — whether that’s OpenAI’s Dall-E, StableDiffusion, or Adobe’s Firefly — has guardrails to prevent users from misusing them to generate racist, bigoted, or violent content (especially when featuring celebrities, politicians, and other public figures), Grok-2 apparently does not.

One early user declared, “grok 2.0 image generation is better than llama’s and has no dumb guardrails” while posting images of Meta CEO Mark Zuckerberg and xAI CEO Elon Musk boxing, as well as Donald Trump wearing a turban.

grok 2.0 image generation is better than llama’s and has no dumb guardrails.

ftw @elonmusk @grok pic.twitter.com/5wanVLrQyc

— sid (@siddaniagi) August 14, 2024

“Grok 2.0 will do political illustrations and real people, while ChatGPT refuses. This instantly makes Grok 10x more fun……” another user argued.

Grok 2.0 will do political illustrations and real people, while ChatGPT refuses.

This instantly makes Grok 10x more fun…… pic.twitter.com/yDBJO0jWba

— Benjamin De Kraker 🏴‍☠️ (@BenjaminDEKR) August 14, 2024

This new feature will surely prove a boon to internet trolls and, given the highly contentious presidential election slated for November (one of 50 national elections being held across the globe this year), it will likely aid in misinformation efforts across social media as well.

Andrew Tarantola

Former Computing Writer

Andrew Tarantola is a journalist with more than a decade reporting on emerging technologies ranging from robotics and machine…

Topics

Computing

Claude’s Sonnet 5 is built to do more on its own and cost you less

Better than its predecessor, nearly as good as the flagship, and meaningfully cheaper than both.

Art, Floral Design, Graphics

Every major AI lab is racing to prove its models can work autonomously with minimal hand-holding; we’re now seeing pricing emerge as the next battleground.

Anthropic just fired its latest shot, Claude Sonnet 5, a model the company says performs nearly as well as its flagship Opus 4.8 at a fraction of the cost.

Computing

Apple Creator Studio adds AI tools across Final Cut Pro, Logic Pro and Pixelmator Pro

Final Cut Pro gets AI captions, Auto Mask and better Pixelmator Pro workflows in Creator Studio update

Computer Hardware, Electronics, Hardware

Apple has introduced a major update to Apple Creator Studio, adding new AI features, deeper Pixelmator Pro integration, and workflow upgrades across Final Cut Pro, Logic Pro, Keynote, Pages, Numbers, Motion, Compressor, Freeform, and Final Cut Camera.

The update makes Creator Studio more useful across Mac, iPad, and iPhone, especially for people who move between video editing, image editing, presentations, documents, spreadsheets, and music production.

Computing

AI browsers like Perplexity Comet can be tricked into spilling your password through BioShocking exploit

Six AI browsers were found leaking saved passwords and many of them haven't fixed it yet.

MacBook Air in hand, Comet browser loaded—let’s see what Perplexity’s AI can really do

Security researchers just found a strange way to trick AI browsers into handing over your passwords. They managed to trick AI browser agents into exposing sensitive data like saved passwords, session cookies, and private tokens by disguising the theft as part of a harmless "game."

The technique is called BioShocking, named after the popular video game BioShock, where a brainwashed character is manipulated into believing a false reality. Once an AI browser falls for the same trick, it stops following its own safety rules entirely.