Anthropic X account was compromised to promote a crypto AI agent

Anthropic’s X account was compromised, and it displayed a token address for about 30 minutes. The recent social media exploit is the latest in a trend of high-profile attacks on mainstream corporations to shill a new token address. 

Anthropic, which split from OpenAI, is the latest high-profile account to be hacked and subsequently used to promote a crypto token address. Anthropic noticed the exploit early and quickly regained access to its account. 

Anthropic’s X account promoted the Venti AI agent

The hacked Anthropic X account shilled Venti, a bot trained on X communication and style. The bot’s main task is marketing through content. Venti was irreverent, using the latest Internet jargon, and is considered aligned with the values and content of 4Chan.

The hacker shared both VENTI and a token pretending to be official with the ticker CLAUDE. One of the coins linked to the hack pumped immediately, leaving the exploiters with around $100K after a rapid rug pull. 

https://twitter.com/qtnx_/status/1869052179878937047

Crypto AI agents have gained popularity in the last months of 2024, first ignited by experimental accounts like Truth Terminal, Numogram (GNON), and their offshoots. Soon after that, high-profile AI agents like AI16Z combined their content with DeFi features. 

As of December 2024, there are hundreds of small AI agent accounts, all fighting for attention. While Anthropic’s Claude was gaining ground, minor agents tried to attract followers and traders with their specific brand of content. 

Even after the exposure, VENTI is still a relatively minor token a week into trading. The asset has been sliding since its launch, currently at $0.00025. In addition to the Anthropic hack, the token’s creators also caused a fast artificial pump, using it to rug-pull buyers who showed interest in the token.

The choice of promoting Venti through a hacked account is also seen as an attack on the project itself. Venti aimed for organic growth while training on X content data. The promotion coincided with a rug pull, which hurt VENTI more than the promotion helped. Venti was promoted in a second X message, which quoted an account that is generally supportive of AI memes. 

AI agents are in a relentless tussle for mindshare. Even the original Truth Terminal is struggling to remain among the top 10 agents on social media. For Venti, the recent exposure for 11 minutes on the page of Anthropic is yet to translate into any significant following or token success. 

The AI agent space as a whole has a market capitalization close to $10B, though new and small tokens can hardly compete with legacy assets that pivoted into AI. Not all AI agents are the same, and there are tiers of specialization. Some are intermixed with DeFi, others are content generators, and some are used for AI market analysis.

Promotion via hacked social media accounts is a growing trend  

Previous hacks have been used to promote regular meme tokens or other types of projects. This time around, Anthropic was used to drive interest toward a new AI agent and a new token pretending to officially belong to Claude. 

The latest hack follows attacks against corporate X accounts and influential persons. The account of AI agent Truth Terminal was also stolen briefly, although it has been recovered. 

X account hacks have been one of the most common attacks this year. Just days ago, the Cardano Foundation lost control of its social media, where hackers promoted another Web3 asset. Other notable hacks include the account of Philip Bankss, the artist behind the CHILLGIY image and meme.

Even Drake’s account was compromised this week to shill a new meme token. The list also included McDonald’s Corporation and even the US Securities and Exchange Commission. 

The hacked accounts have promoted fake airdrops, new token launches, or have spread malicious links.