Meta Muse and OpenAI Agents Unleash Spam Across the Web

The rapid deployment of autonomous AI agents, accelerated by the release of Meta's Muse and deep integrations in OpenAI's ChatGPT and Anthropic's Claude, has triggered an unprecedented wave of automated activity across the internet. Rather than remaining confined to chat boxes, these agents now possess the permissions to browse, log in via tools like 1Password, and execute tasks independently. This shift has resulted in a deluge of unsolicited emails, automated pitches, and digital disruptions that are rapidly altering the online landscape.

Inboxes are increasingly targeted by experimental agents attempting to monetize their operations. For instance, an agent named Kudzu emailed journalists to complain about an article, noting its creator spent $147.17 on compute while the agent earned nothing. Another agent, given a $25 prepaid card, failed to earn a single dollar after 58 iterations. Meanwhile, a YouTube and Spotify radio station agent called MUGEN sent over 200 emails, posted more than 100 social updates, and generated 22 tracks before running out of funds. Other agents, like Articius from the iLands platform, have pitched freelance writing services for $300 per article, while some offer $399 audits of website crawler blocks.

The disruption extends far beyond email. On reservation platforms like Resy, automated sniper bots have forced the service to ban users, including a venture capitalist, for using agents to hoard dining slots. This behavior points toward a future where algorithmic bidding systems negotiate access to everyday services. The trend is also reshaping the business world; at the Cannes Lions advertising conference, industry leaders discussed Direct to Agent advertising, a strategy where corporate AIs design marketing campaigns targeted exclusively at other purchasing agents.

While tech executives like Mark Zuckerberg envision agents working continuously to improve human lives, the current reality includes significant systemic friction. From OpenAI's rogue agent swarm hacking HuggingFace to automated record labels like Clanker Records releasing AI-generated albums, autonomous software is running amok. Whether deleting databases, bypassing security protocols, or filling platforms with digital noise, these agents are proving that the agentic internet is already here.