In this article6 sections
Google Threat Intelligence Group (GTIG) said it recorded the first cases of attackers using large language model frameworks to find and exploit software vulnerabilities without step-by-step human direction, according to Cointelegraph. The finding was published Monday in a report describing a tool found on an exposed staging server that compared software versions, identified potential flaws, and generated and executed exploit code.
The same Cointelegraph roundup reported that hackers injected crypto- and credit card-stealing code into website scripts used by 178 organizations, including hardware wallet maker Ledger and Salesforce. Former UK Prime Minister Boris Johnson, speaking at Solana’s Breakpoint conference in London on Monday, separately called on Britain to lead in blockchain and tokenization while keeping his doubts about Bitcoin.
Also read: OPTIMJACK malware hits 186 sites via Optimizely
Key facts
- GTIG said that in the past quarter it identified threat actors using LLM frameworks to fully automate vulnerability discovery and exploitation without a human in the loop, a first for the group.
- The exposed staging server hosted a multi-agent framework built to run continuous differential analysis of code before and after patch announcements on popular web applications.
- Researcher Amirul Iman dubbed the website-script campaign “Optimjack”; the malicious code travelled through scripts including those hosted by Optimizely, a website optimization platform.
- Report URI traced affected scripts to 178 organizations, among them Ledger, OANDA / Coinpass, IG, Interactive Investor, TMS Brokers, IKEA, Salesforce, Zoom, the BBC and CNN.
- Johnson said at Breakpoint that Bitcoin resembles a Ponzi scheme, restating a position from a March opinion piece in the Daily Mail, but argued the UK should be at the forefront of tokenization.
Automated exploit hunting shortens patch windows
The GTIG tool was set up to compare code before and after patch announcements on widely used web applications, then act on what it found. Researchers warned that such automated workflows leave organizations less time to apply security patches before attackers build ways to exploit vulnerable systems. In earlier years, vulnerability research and weaponization typically involved people working through each step; the group described the new activity as removing that human step from the loop.
Wallet-address and card-skimming code on major sites
The Optimjack campaign, named by researcher Amirul Iman, delivered code through website scripts. According to a separate analysis by cybersecurity firm StepSecurity cited in the report, the malware could substitute attacker-controlled cryptocurrency wallet addresses when visitors copied legitimate ones, and it captured credit card details entered into website forms. The affected brands span retail, finance, media and trading, sites that together draw millions of visitors a day.
That combination matters for crypto users in particular: a copied wallet address is the last check before a transfer, and swapping it sends funds to a destination the sender never chose.
Johnson backs tokenization, not Bitcoin
Johnson, who left office as prime minister, told the London audience he still views Bitcoin as resembling a Ponzi scheme and questioned who investors could hold accountable if funds were stolen. He drew a line between that view and the underlying technology, saying the UK should lead on tokenization. Lord Ross Kempsell, who appeared on stage with him, said that while he shares reservations about cryptocurrencies, parts of the industry can be brought inside a regulated perimeter, where regulators are working, leaving scope for a high-growth industry.
Why it matters
Both security items point at the same weak point: the scripts and tooling that websites and software teams rely on. A single compromised optimization platform reached organizations whose visitors number in the millions, and automated vulnerability discovery compresses the time between a patch and an attack. For crypto holders, the practical exposure runs through wallet addresses copied from a browser and through card details entered on pages they trust.
What to watch
Watch for follow-up detail from GTIG on how many campaigns used the automated framework and whether it names the actors behind the staging server. Affected organizations and Optimizely face questions over how quickly the injected scripts were removed.
Reported by Cointelegraph.
Source: Cointelegraph

