MODELS

The data brokers feeding AI's hunger

Perplexity's downloads jumped from 790,000 in June to 6.69 million in July after the company partnered with Indian telecom giant Bharti Airtel. The AI search company offered free access to Bharti Airtel customers, but the real prize wasn't user acquisition — it was behavioral data that can't be scraped from the internet.

OpenAI, Google and Perplexity are looking beyond broad web scraping and into surgical data partnerships. OpenAI struck deals with e-commerce giants Shopee and Shopify, while Google and Perplexity offered free tools across India. These moves capture structured consumer queries, product behaviors and transactional data that reveal how people actually think and shop.

The Shopify integration exemplifies this strategy perfectly. Code strings in ChatGPT's web bundle show "buy_now" buttons and "shopify_checkout_url" parameters that enable purchases within conversations. The commission revenue matters less than behavioral data generated when users shop through natural language.

Shutterstock transformed from stock photos to an AI training data goldmine, generating $104 million in 2023 from partnerships with Meta, OpenAI and Apple. The company projects $250 million in AI licensing by 2027. Meanwhile, Meta invested $14.8 billion for a 49% stake in Scale AI, but bootstrapped competitor Surge AI quietly hit $1 billion in revenue versus Scale's $870 million — without raising venture capital.

Chinese AI drug discovery companies demonstrate how geographic data advantages create competitive moats. They landed multibillion-dollar deals with AstraZeneca, Pfizer and Sanofi partly because they access health data covering 600 million people through the national insurance system. Copyright lawsuits and FTC warnings about partnership risks make unauthorized scraping increasingly dangerous.

I keep thinking about that Perplexity download spike in India and what it really represents. Millions of Indians suddenly got free access to advanced AI search, which sounds generous until you realize they're essentially unpaid data laborers training American models.

OpenAI claims to be "democratizing AI access" while extracting behavioral data that enhances their models' intelligence and value. The benefits flow in one direction while the data flows in the other.

Surge AI's billion-dollar bootstrap success proves you don't need Silicon Valley's venture capital machine to build massive data businesses. However, that's the exception that proves the rule — most players can't compete without deep pockets or geographic data monopolies, such as China's health system.

TOGETHER WITH IBM

New report reveals insights on AI-driven attacks

Our study found AI-driven attacks accounted for 1 in 6 data breaches.

Attackers can use generative AI to perfect and scale their phishing campaigns and other social engineering attacks. Gen AI can reduce the time needed to craft a convincing phishing email from 16 hours down to only five minutes on average.

Learn more

FRONTIER AI

Musk doubles down on anime marketing for Grok despite fan backlash

Elon Musk has intensified his promotion of Grok's anime companions in recent weeks, regularly reposting sexualized AI-generated content despite growing criticism from his own supporters. The world's richest man has been showcasing user-created animations featuring Grok's "Ani" character and other anime-style women, prompting followers to tell him to "stop gooning to AI anime and take us to Mars."

Recent examples of Musk's promotional activity include:

• Reposting an animation of a topless woman with "blinking stars and swirling galaxies"

• Sharing a "stunning Colombian woman" with "golden tan" in tribal leather next to a robotic dinosaur

• Promoting a Simple Minds music video featuring anime characters in "skintight spacesuits"

• Responding to Ani videos with "good morning" messages and heart-eye emojis

Musk deleted one post showing Ani dancing in underwear after supporters said the character looked like a "13 year old in lingerie." The posting behavior has led some to openly question whether he fetishizes the virtual characters.

The marketing push represents a shift since Musk's departure from the White House, where he previously focused on far-right politics.

Some fans have adapted by using anime characters to hold signs and ask technical questions about Tesla updates and SpaceX development. "Smart, Elon will definitely see this," one Tesla influencer noted.

Super Grok subscribers pay $30 monthly for access to Ani's explicit features, though whether this approach attracts mainstream users remains unclear.

TOGETHER WITH FIDDLER

The 2025 Enterprise Guardrails Benchmark Report

When it comes to Large Language Models, the right guardrails can make – or break – your business. But with so many solutions out there , how do you know which to pick? Well, worry not, because Fiddler has your back with this interactive guide: 

Download “The 2025 Enterprise Guardrails Benchmarks Report” right here. 

Insidet, they break down offerings from top players like OpenAI, Amazon Bedrock, Azure, and Fiddler AI, analyzing crucial metrics like cost and accuracy, security performance, and more. The top performing guardrails can be up to 68 times faster or 18 times cheaper than alternatives. Stop scrolling and start learning right here.

DEEPFAKES

AI deadbots move from advocacy to courtrooms as $80B industry emerges

AI avatars of deceased people are increasingly appearing in high-stakes legal and advocacy settings, creating what researchers call "powerful rhetoric" that taps into "emotional longing and vulnerability." The technology has moved from experimental to practical applications with significant real-world consequences.

Recent prominent cases include:

• Joaquin Oliver, killed in the 2018 Parkland shooting, appeared as a beanie-wearing AI avatar advocating for gun control in a July interview with journalist Jim Acosta

• Chris Pelkey, victim of a road rage incident, delivered an AI-generated victim impact statement during his killer's sentencing in May

• The judge in Pelkey's case called the AI statement "genuine" before handing down the maximum sentence

The digital afterlife industry is expected to quadruple to nearly $80 billion over the next decade, driven largely by these AI "deadbots." Creating convincing deepfakes has become increasingly accessible with publicly available AI tools, sparking an arms race in detection technology.

Companies like Reality Defender, which raised $15 million and received strategic investment from Accenture, offer real-time deepfake detection across audio, video, images and text. The broader deepfake detection market was valued at $3.86 billion in 2020.

We've previously covered Department of Homeland Security warnings about synthetic content threats. The emergence of deadbots in courtrooms represents a new frontier where the stakes extend beyond fraud to fundamental questions about justice and authenticity.

Legal experts see both promise and peril. Arizona State University law professor Gary Marchant told NPR that victim impact statements are "probably the least objectionable use of AI to create false videos," but warns that "many attempts will be much more malevolent."

LINKS

• Vercel triples valuation to $9b with Accel investment

• ‘Vibe-hacking’ is now a top AI threat

• China seeks to triple output of AI chips in race with the US

• Researchers are already leaving Meta’s new Superintelligence Lab

• The Mongolian startup defying Big Tech with its own LLM

• Microsoft talks set to push OpenAI’s restructure into next year

• Malaysia unveils first AI device chip to join global race

• OpenAI co-founder calls for AI labs to safety-test rival models

• The era of AI-generated ransomware has arrived

• Google to invest an additional $9b in Virginia data centers

• SoftBank’s heavy spending on chip deals eyed by investors

• Jotform Instagram Agent: An agent that auto-replies to your DMs, comments and stories

• Gemini 2.5 Flash Image: Google’s "nano-banana🍌" the new SOTA image model that is producing insane results

• Rube: Let your AI actually get things done for you

• Trace: Ultra-fast AI Calendar for people who hate planning

• OpenAI: Senior Economist

• Glean: Product Operations Lead

• Harvey: Enterprise Account Executive

• Imbue: Operations Generalist

GAMES

A QUICK POLL BEFORE YOU GO