Sponsored by
Hello readers,
Welcome to the AI For All newsletter! Today, we’re talking about AI taking scams to the scammers, the biggest news of the week, and more!
AI in Action: Learning to play the perfect victim
An Australian startup called Apate.ai has built a business on getting scammed on purpose. As TechCrunch reported this week, the company runs thousands of AI agents posing as ordinary people of different ages, genders, accents, and languages. They answer scam calls, reply to sketchy texts and emails, click phishing links, and join the social media groups where fraudsters go fishing. Apate says the fleet has burned through 17 years of scammers’ time across 2.5 million conversations. The wasted hours are a side benefit. The product Apate sells to banks, telcos, and government agencies is what scammers let slip while they think they’re closing a mark: bank accounts, crypto wallets, phone numbers, and the brands they’re impersonating.
The idea started at a family picnic. Several years ago, founder and CEO Dali Kaafar got a scam call, played dumb for his kids’ amusement, and kept the caller on the line for 44 minutes, time that scammer didn’t spend on a real victim. That turned into research at Macquarie University, where Kaafar’s team trained chatbots on recorded scam calls, emails, and social media chat logs. Apate now runs a mix of frontier models and its own fine-tuned ones, trained partly on its agents’ past conversations with scammers, and customers can hand the agents specific missions, like extracting a particular wallet address. When two WIRED reporters tested one of the bots by tag-teaming it with a crypto pitch, they spent six minutes trying and couldn’t get it to invest. Apate sends AI agents undercover to get scammed so they can gather fraud intelligence.
Most fraud defense kicks in after contact. As co-founder Brad Joffe put it to TechCrunch, telcos block numbers that scammers simply replace, and banks flag transactions after the scammer has already reached the customer. Apate’s approach collects that intelligence before anyone loses money. Its customers include Commonwealth Bank of Australia and TPG Telecom, and the company says TPG has diverted nearly 590,000 calls to its agents, while another Australian bank used them to disrupt 22 scam campaigns. Those figures are Apate’s own. The pressure behind them is well documented, though: the FBI received more than 22,000 complaints of AI-related crime in 2025, with about $893 million in losses. Apate even tracks how often scammers swear at its bots, and celebrates every one.
🔥 Rapid Fire
Commentary: It’s Time To Stop Using ARR
Analysis: The Hater’s Guide to Junk Bonds
Character AI files for bankruptcy over costs and litigation
Bankruptcy documents show AI’s economics don’t work
CoreWeave stock plunges as bonds turn distressed
80% of neoclouds could cease to exist by 2030 per Gartner
“By 2030, Gartner predicts that cost pressure and a lack of a differentiating business model will force over 80% of neocloud providers to cease independent operations.”
“Through 2030, AI data center owners would need to generate a cumulative $8.5 trillion in revenue to cover announced capital spending, financing costs, and a 12% return on invested capital (ROIC).”
“Based on Gartner projections, a continued AI infrastructure investment trajectory could leave more than $4 trillion in asset value exposed to write-down risk through 2032.”
Jefferies warns of massive capital destruction in AI
BIS report: circular relationships among AI firms
Broadcom CEO likens AI investors to casino gamblers
“A lot of investors have no clue what they are investing in.”
Data center debt deals face increasing scrutiny
Further reading: AI’s Credit Crunch
Anthropic’s revenue growth is slowing down
Investors temper IPO valuation expectations over slowdown
Anthropic said it was profitable on an “adjusted basis”
That means they’re not profitable
5-7% of revenue came from distillation of its models
AI token spend is flat while usage rises
“The growth in new AI customers has slowed down significantly. It feels like we’re at the top of the S-curve, where the growth in the market will have to come from more intense adoption rather than more customers.”
“And this slowdown has been unusually stubborn. Because of August and September, we’ve recorded two straight months of virtually no growth in token spend.”
Accenture employees can now spend only $29 per month on Claude
Are we in an AI bubble?
Which GTM Work Should Agents Own?
Automating at random isn’t a system. Deciding what agents own and what people own is.
On October 28, HubSpot and Dust run Build, Buy, or Hire? live. You’ll leave knowing what to hand to agents, and with the workflows to prove it:
A framework for where a human stays in the loop
A post-call automation that updates your CRM for you
A Slack view of who owns each account and what happens next
Live at 11am EDT / 8am PDT
📖 What We’re Reading
For decades, getting a professional manicure has been a highly manual experience. A customer typically visits a salon, chooses a color or design, and relies on a technician to apply it with enough precision to produce a clean result. That process is now becoming an interesting target for automation.
