DeeperML

xAI, led by Elon Musk, has launched Grok-3, the latest iteration of its large language model, boasting advanced reasoning capabilities. The company claims that Grok-3 represents a substantial advancement over its predecessor, Grok-2, due to its "reasoning models" that mimic human thinking. Researchers added it was trained using significantly more compute power, and early testing suggests it may outperform models from OpenAI and DeepSeek. Grok-3 features both a large complex model (Grok-3 Reasoning beta) and a small fast model (Grok-3 mini Reasoning) for generating responses to complex questions.

To access the reasoning capabilities, users can engage "Think" mode and "Big Brain" mode for more complicated queries. Grok-3 can be paired with "DeepSearch," to scan the internet for relevant knowledge to incorporate into its answers. However, the launch has not been without controversy, as reports have surfaced that Grok-3's search results are being censored. Users have reported that Grok-3 is ignoring sources identifying Elon Musk or Donald Trump as misinformation spreaders. Additionally, Grok-3 is highly vulnerable to indirect prompt injection, where malicious instructions can be inserted into tweets, causing Grok to misbehave in response to certain queries.

ImgSrc: d15shllkswkct0.

References :

• SiliconANGLE: Elon Musk’s xAI unveils Grok-3 with advanced reasoning capabilities
• THE DECODER: X chatbot Grok 3 rebels against Musk and Trump, gets its search results censored
• TestingCatalog: xAI to add file attachment support in DeepSearch mode for Grok 3 on X

Classification:

• HashTags: #Grok3 #LLM #AISocialMedia
• Company: xAI
• Target: X users
• Attacker: xAI
• Product: Grok-3
• Feature: AI Chatbot
• Malware: Prompt Injection
• Type: AI
• Severity: Informative