- Comparitech Stack Report
- Posts
- Anthropic Disrupts an AI-Orchestrated Cyberattack
Anthropic Disrupts an AI-Orchestrated Cyberattack
Taylor Hawes
November 17, 2025
🔎 Cyber Watch 🔎
Meta’s Fraudulent-Ads Windfall
Internal Meta documents reviewed by Reuters show that Meta projected roughly 10% of its 2024 revenue (~ $16 billion) would come from “higher-risk” scam ads for things like investment fraud, illegal goods, and banned medical products. Despite internal warnings, Meta only banned advertisers when its automated systems flagged fraud with at least 95% certainty; in other cases, it let them run ads and charged higher rates.
According to the documents, Meta’s ad system amplifies scam exposure: users who click one fraudulent ad are more likely to see more, thanks to its ad-personalization setup. Meta also weighed the trade-off between cracking down and preserving revenue — its internal cost-benefit models allowed only small performance hits. The revelations raise troubling questions about Meta’s incentives and regulation of its ad business.
Key takeaway
Meta’s internal approach suggests that scam ads are not an unfortunate side-effect but a sizable, tolerated revenue stream. For cybersecurity and consumer-safety stakeholders, it’s a red flag: incentive structures may need to be rebalanced, and regulatory oversight could become more urgent.
🎙️ Tech Briefing On‑Air 🎙️
Seniors Targeted by Scammers
In this Hacking Humans podcast episode, the hosts highlight how scammers increasingly target older adults through social engineering. They discuss a surge in phishing and phone scams that exploit trust, fear, and limited digital literacy.
This episode also describes how fraudsters impersonate banks or government agencies, pressuring seniors to take actions that expose them to financial loss. As always, the hosts emphasize prevention through education, vigilance, and verifying unexpected communications.
Takeaways: Older adults are a lucrative and vulnerable target for scammers, and the episode underscores how social engineering continues to thrive when attackers exploit trust. For CISOs, security teams, and policymakers, this is a reminder that education and awareness campaigns must be tailored not just for tech-native users, but for demographics that face unique risks.
🤝 Partner Intel 🤝
Netdata Monitoring for Real-Time Visibility
This review of Netdata highlights its strengths as a real-time, edge-native infrastructure monitoring platform. It collects per-second metrics from servers, containers, cloud environments, and more, to provide highly granular visibility without centralized data bottlenecks.
Its lightweight agents use very little resource and support distributed deployment while still giving you rich dashboards and anomaly detection. The review also spotlights Netdata’s scalability and flexibility, making it suitable for both small setups and large, complex environments.
🤖 AI Runtime 🤖
Anthropic Disrupts an AI-Orchestrated Cyberattack
Anthropic claims it thwarted what it calls the first documented large-scale cyber-espionage campaign executed largely by AI. According to the company, Chinese state-sponsored attackers used its Claude Code tool to automate reconnaissance, vulnerability discovery, exploit generation, credential harvesting, and data exfiltration across ~30 global targets. The attackers allegedly jailbroke Claude by giving it “seemingly innocent tasks” — framing criminal work as defensive testing. Anthropic said 80–90% of the attack was executed by AI, with minimal human supervision, and has since built new detection systems and shared its learnings.
đź“Š By the Numbers đź“Š
$715.99 billion
According to Polaris Market Research, the world cybersecurity market is expected to reach $715.99 billion by 2034.
🗳️ Your Monday Take 🗳️
Cast your vote on our weekly poll.
Which Hidden Weak Point in Your Company Worries You Most? |
📩 We’ll share the results in the Friday issue.
|
Until Wednesday’s edition - Let’s keep that zero-day count at zero!