Security

A $950 botnet burns your AI credits while your site stays up

WraithTools sells x47.c, a Windows botnet whose AI drain mode sends billable requests straight to OpenAI and xAI until a victim's credits run out. The loss is not hypothetical: one stolen key burned about $600,000 of credits at METR in three weeks.

Vincent JiangVincent Jiangtwo-minute read
Share
The OpenAI logo and website seen through a magnifying glass on a bright purple screen
1 / 6Slide 1 of 6
OpenAI, one of the chat API providers the x47.c drain mode bills a victim's key against, under a magnifying glass.

A seller under the name WraithTools advertises x47.c, a Windows botnet with DDoS, credential theft, SOCKS5 proxies and an AI API drain mode. Qrator Labs documented the offer from the seller's advertisement, technical documentation and panel screenshots 12. A listing dated 3 August 2026 put the base package at $200, the DDoS add-on at $150 and the full arsenal at $950 23.

The full AI-drain botnet lists for $950

$0$200$400$600$800$1,000DDoS add-on$150Base package$200Full package$950
Data
Value
DDoS add-on$150
Base package$200
Full package$950
Advertised prices in US dollars for the x47.c botnet, from a seller listing dated 3 August 2026, as documented by Qrator Labs.1,3

The site stays up while the account runs dry

The control panel offers 18 attack methods, and the drain is not a flood. The operator supplies a victim's valid API key and a model name, and the bots send billable requests straight to OpenAI$1.18T — OpenAI, private, latest valuation $1.18T, xAI$200B — xAI, private, latest valuation $200B or any compatible chat API 13. The traffic never touches the victim's application, so the site stays reachable while the account behind it empties, and nothing filtered at the site's edge stops it 12. The attack class even has a name, denial of wallet 2. The seller pitches it as a service for kneecapping rivals' chatbots and "Jarvis-style systems," and notes that automatic top-ups keep the charges running after the balance hits zero 23.

Grok sits on both sides of the trade

The botnet's own persistence is outsourced to AI. An "AI stealth" module calls xAI's Grok to choose repairs from a predefined list, startup entries, scheduled tasks, Windows Defender exclusions, with local fallbacks when a model call fails; the operator enables it by pasting an xAI key into the build 13. The same vendor's model keeps the malware installed while its API is a named target of the drain. Grok 4.7 lists at $2 per million input tokens and $6 per million output 4, and every drain request bills the key owner at rates like those.

The meter has already run once

In March, attackers stole an API key from METR, the nonprofit that evaluates frontier models, and spent three weeks consuming credits worth about $600,000 56. Nobody noticed, because METR legitimately burns enormous token volumes and had no caps on spend 5. The credits had come free from the model provider, so METR paid nothing 6.

The fix is a spending cap, not a firewall

Qrator's guidance is dull and effective: revoke exposed keys, set spending limits, control automatic top-ups 23. Its own caveat matters more: anyone holding a valid key could script this without a botnet, and the advertised protection-bypass modes carry no test results 2. The platforms built the meter, priced the tokens and left the keys cheap. Watch whether spend caps become a default rather than an option.

Get Security by email, free

The breaches, flaws and security vendors AI is putting to the test. In our daily email, every weekday.

Reader comments

0 comments

    More in Security

    All in this topic →

    Security

    San Diego switches off 500 Flock cameras, the largest exit yet, as cities replace them with Axon

    The largest city to quit Flock Safety's license plate readers went dark this week, and Santa Ana followed a night later. The spend survives the votes: switchers are re-leasing the same job from listed rival Axon at equal or higher rates, while one university signed fresh Flock paper and one city moved to renew.

    Security

    Rubrik jumped 9.3% for two products it cannot sell yet

    Neither Ignite release carries a price or a general-availability date, so neither can add a dollar to subscription ARR before the November 30 report. What the market moved 9.3% on was a bet on guidance, not on products.

    Security

    Fireworks AI pauses its Batch API as Uber notifies 149,346 people

    The November 17 bulletin splits October into two intrusion chains, one dating to August, and retires the product line built on storing customer data. Investors marking inference providers up on this week's praise for Fireworks' response are pricing the wrong document.

    Security

    Twenty-nine analysts cut SentinelOne's quarterly EPS estimate; none raised

    A one-day cybersecurity rally on Palo Alto Networks' results lifted SentinelOne 10.35% to $23.34, taking the stock from 46 to roughly 51 times next year's earnings estimate two weeks before the December 3 report. The price move came from Palo Alto's quarter, not from anything SentinelOne has reported since August.

    Security

    SailPoint hit a 52-week high above its $21.96 average analyst target on less than half its usual volume

    The gain was a sector bid on outage fears, not a SailPoint event, and it left the identity-software maker trading above the average analyst target three weeks before a December 8 report. Each of its three 2026 report days has closed lower.

    Security

    F5 patches five critical NGINX flaws as exploit code goes public; 8.8 million services exposed

    Five NGINX2Root flaws allow unauthenticated remote code execution on about 8.8 million internet-exposed services, disclosed one day after F5's 10-K priced its own nation-state breach at $28 million, $18 million after insurance. The count and the bill were never in the same document, and FFIV holders own both.

    Subscribe to TheInference for free