AI Agents Are Quietly Hiring, Hunting Threats and Stealing Work

AI agents are quietly graduating from answer boxes to autonomous workers that shop, hire and even patrol corporate networks — and that shift is already reshaping who gets paid and who gets conned.

On the streets, UK startup MeatLayer pitches itself as “the world’s first human marketplace for AI agents,” where software agents post hyper‑granular gigs — snapping photos, moving packages, verifying locations — and pay workers directly via Stripe once proof is uploaded, with advertised top earners making thousands of dollars a week. Tasks on MeatLayer are surfaced to nearby users and explicitly framed as being “posted by AI agents running for businesses and individuals,” creating a gig economy where workers report not to managers but to code. MeatLayer positions this as frictionless income; labor advocates see a template for algorithmic bosses with even less accountability than today’s apps.

In security, that same agentic model is hunting threats. CrowdStrike’s new “Agentic Security Workforce” embeds AI agents that continuously correlate telemetry, proactively search for adversaries and can trigger remediation across the “kill chain,” aiming to free human analysts for higher‑level judgment while software handles detection, triage and response loops. CrowdStrike is not alone: IBM and others now describe autonomous security operations centers where coordinated agents shoulder most of the monitoring and incident workflow with “minimal human involvement.” IBM That promises faster containment — and raises the stakes if an attacker hijacks the agents themselves.

To make any of this trustworthy, identity and memory are becoming infrastructure. Worldcoin’s parent, World, is expanding its World ID “proof of personhood” so digital identities can be bound to agents, and has floated linking AI agents to verified humans on its new World Chain network. TechCrunch Meanwhile Memories.ai is selling a large visual memory model for wearables and robots that lets systems recall faces, places and past interactions, pushing agents from session‑based tools toward continuous, lived context. Tech Times As agents gain wallets, long‑term memory and real‑world labor, the line between “assistant” and autonomous economic actor is dissolving — and with it, assumptions about who is really on the other side of any interaction.