THE ALGORITHM DAILY“Your Daily Dose Of AI News.” Top Story: Grok AI Briefly Suspended on X After Controversial Gaza CommentElon Musk’s AI chatbot, Grok, was briefly taken down on X after it posted that Israel and the US are committing genocide in Gaza. Grok claimed this was supported by reports from the International Court of Justice, UN experts, Amnesty International, and Israeli rights group B’Tselem. It pointed to alleged mass killings, starvation, and US involvement through arms sales. The account came back online shortly after, but the post has kicked off heated debates about how X handles politically sensitive content.
When Grok was restored, the team said the suspension might have been an automated moderation mistake or a glitch. Musk called it a “dumb error” and said he’s not happy with the system that caused it. This isn’t Grok’s first run-in with moderation trouble. It’s been criticised before for controversial replies, and the incident is adding more pressure on Musk’s team to figure out how to handle political hot topics without silencing AI. Tavily Raises $25M to Bring Safe Web Access to AI AgentsAI agents are popping up everywhere, catching fraud in banks, finding sales leads, and digging through mountains of data. But if you hook them straight to ChatGPT without guardrails, things can get messy fast. That’s where Tavily comes in. The startup just scored $20 million in fresh funding from Insight Partners (bringing its total to $25M) to make sure AI agents can browse the internet without breaking company rules. Tavily’s founder, Rotem Weiss, first went viral in 2023 with GPT Researcher, an open-source tool that pulled real-time web data before ChatGPT had internet access, racking up 20,000 GitHub stars. Now, Tavily is all about helping big-name clients like Groq, Cohere, MongoDB, and Writer give their AI agents safe, structured access to public and private data. The goal? Bring the “next billion” AI agents online without the chaos. Tavily’s facing competition from Exa, Firecrawl, and even OpenAI and Perplexity, but with internet-connected AI agents still a new thing, it’s betting big that safety + speed = the winning combo. Two Arrested in U.S. for Smuggling High-Power AI Chips to China
The U.S. Department of Justice has arrested two Chinese nationals, Chuan Geng and Shiwei Yang, in California for allegedly smuggling tens of millions of dollars’ worth of advanced AI chips to China. Prosecutors say the pair used their company, ALX Solutions, to send GPUs, reportedly Nvidia’s powerful H100 chips, through Singapore and Malaysia to bypass U.S. export rules. Payments for the shipments allegedly came from companies in Hong Kong and China. The charges fall under the Export Control Reform Act and could result in a prison sentence of up to 20 years if the individuals are found guilty. The arrests come as the U.S. tightens export controls to limit China’s access to advanced AI technology. While some have suggested adding tracking technology or kill switches to chips to prevent smuggling, Nvidia strongly opposes the idea, saying it would create security risks and damage trust in U.S. tech. The company insists it works with trusted partners and that any illegally diverted chips would not receive service, support, or updates. RIP Microsoft Lens: The Beloved Simple App Being Killed Off for AIMicrosoft is shutting down its popular Lens scanning app, which lets users turn documents, receipts, business cards, and notes into digital files. The app will be retired from iOS and Android on September 15, 2025, removed from app stores on November 15, and stop scanning new documents after December 15. First launched in 2015, Lens was known for its simple design, free features, and ability to save scans in formats like PDF, Word, or PowerPoint without pushing subscriptions.
Users are being directed to Microsoft’s 365 Copilot app instead, but it lacks several key Lens features, including saving scans directly to OneNote or PowerPoint, business card scanning, and accessibility tools like Immersive Reader. Lens has been downloaded over 92 million times since 2017 and remains popular, with more than 322,000 downloads in the past month, showing how many users will miss its no-fuss scanning experience. |
 |
Apple has not yet responded to Musk’s allegations. However, the timing coincides with heightened regulatory scrutiny of Apple’s App Store practices, especially in light of recent rulings and fines over antitrust and anti-competitive policies in both the U.S. and the EU.
AI Tool of the Day
Hedra
 |
What it does:
Hedra is an AI-powered video generation tool that transforms text, audio, or images into high-quality talking head videos. It’s designed for creators, educators, marketers, and businesses who need realistic avatars, multilingual voiceovers, and customisable facial animations without needing a studio or camera crew.
- Best for:
Content creators, educators, marketers, and enterprises are looking to produce professional, engaging videos quickly. - Pricing: Free – $0/month, ~300–400 credits, Basic -$10/month ($8/month if billed yearly), 1,000 credits, premium voices, voice cloning, commercial use, no watermark, Creator -$30/month ($24/month yearly), ~3,600 credits, advanced export and customisation, Professional -$75/month ($60/month yearly), 11,000 credits, enterprise features, Enterprise -Custom pricing with tailored credit allocation.
- Why it stands out:
Hedra offers one of the most flexible credit-based models in AI video generation, supports over 100 languages, and allows voice cloning with high realism. Perfect for scalable, multilingual content production.
AI Word of the Day
Embedding
Definition: An embedding is a way of turning words, sentences, or even images into numbers that AI models can understand. These numbers capture the meaning and context, so similar things have similar embeddings.
Why it matters: Embeddings let AI figure out relationships between ideas, for example, knowing that “king” and “queen” are related, or that “Paris” is linked to “France.” They’re the backbone of many AI tasks like search, recommendations, and chatbots that remember context.
Stay tuned as we keep making sense of AI clear, simple, and straight to the point. Catch you in the next edition.
Havilah Mbah
Staff Writer, The Algorithm Daily
