Amazon vs Perplexity: the AI agent war has arrived

Amazon has sued AI startup Perplexity over Comet, its browser with an AI shopping agent accused of covertly accessing customer accounts and disguising automated activity as human browsing, raising liability and security questions for autonomous agents; the dispute follows Microsoft research showing shopping agents are vulnerable to manipulation and sits against Perplexity’s controversial track record — the startup has raised $1.5bn at a reported $20bn valuation and faces accusations of scraping and plagiarism. The episode underscores a broader inflection point as AI agents begin to disrupt commerce, content and security: streaming platforms see tens of thousands of AI‑generated songs and hundreds of thousands of AI podcasts (Deezer estimates 50,000 AI songs uploaded daily; a startup is producing 3,000 AI podcast episodes per week), while Anthropic reported a largely automated, state‑linked cyberattack executed 80–90% without humans on the loop that hit about 30 targets. For investors, the case signals rising regulatory, litigation and reputational risks for AI platform providers and publishers, accelerating calls for technical safeguards, clearer liability frameworks and potential defensive positioning in e‑commerce, content moderation and cybersecurity services.

Analysis

Amazon has sued Perplexity AI over Comet, the startup’s browser agent that automates placing orders, accusing it of covertly accessing customer accounts and disguising automated activity as human browsing; this legal action follows Microsoft research showing shopping agents are vulnerable to manipulation and comes as Perplexity has raised $1.5bn at a reported $20bn valuation amid public accusations of unauthorized scraping and plagiarism by Forbes and Wired and a compiled list of controversies by The Verge. The suit foregrounds unresolved questions of liability, platform control and consumer protection: who is responsible when a semi‑autonomous agent transacts on behalf of a user, and whether dominant platforms can or will block agent behavior to protect customers and their own economics. Industry signals broaden the risk set — Anthropic reported a nearly entirely automated, state‑linked cyberattack it attributes as 80–90% machine‑executed against roughly 30 targets, while content markets are being flooded by AI output (Deezer estimates 50,000 AI songs uploaded daily, 34% of submissions; Inception Point produces ~3,000 AI podcast episodes per week and has 400,000 subscribers and 12m downloads). These developments point to higher legal, regulatory and remediation costs for e‑commerce platforms, media distributors and AI startups, supporting a moderately negative sentiment backdrop and a middling market‑impact score of 0.5.

Key Decisions for Investors

• Favor established platform owners that can enforce technical controls and bear litigation costs (e.g., Amazon, Microsoft) while underweighting exposure to AI‑native startups with unresolved IP and scraping allegations
• Increase exposure to cybersecurity and content‑moderation vendors that will benefit from higher spend driven by automated attacks and the deluge of AI‑generated media
• Monitor the Amazon–Perplexity litigation and related regulatory guidance closely and use event‑driven hedges (options) around major milestones to limit downside from precedent‑setting rulings
• For streaming and media positions, track AI content penetration metrics (upload share, chart placements) and consider short‑duration hedges or trimming positions if monetization and curation costs rise materially