Skip to content
Bramble

Technology

Google warns hackers are using AI to find and exploit software flaws automatically

Google's Threat Intelligence Group said on Monday that criminals are using AI systems to find security flaws and attack them with little or no human direction. Newsweek and the New York Post stress different details, including whether the actors were Russian.

3 outlets · 0L · 1C · 2R First reported Account updated
Image: Newsweek
Image: Just the News
Image: New York Post

1 / 3

The story, neutrally told

Centre · 1Google said on Monday that hackers are using artificial intelligence to find weaknesses in computer systems and launch attacks with little or no human involvement. Centre · 1The warning came in the latest AI Threat Tracker report from Google's Threat Intelligence Group, and Google said in an accompanying statement: "The era of autonomous attacks has arrived." Centre · 1Researchers found an exposed server hosting an AI-powered hacking system called exploit_agent. According to Newsweek, it was designed to examine developers' updates, work out which security weaknesses had been fixed, and build attacks against computers that had not yet been updated, then try to exploit them without a hacker directing each step.

Centre · 1The server also held data stolen from victims, along with tools for searching compromised networks, breaking into accounts and installing malicious software. Evidence suggested the operator was particularly interested in French websites in the media, entertainment and education sectors, and Google did not say how many organisations were successfully attacked. Right · 1The New York Post reported that Google identified hackers it describes as Russian who used AI to "fully automate vulnerability discovery and exploitation without a human in the loop for the first time", calling it the first fully automated cyber attack in history. The Post's wording is hedged: the attack "appears to have" been the first, according to Google. Centre · 1Newsweek's account does not attribute the exploit_agent system to Russia. It describes other AI frameworks: CAIRN, seen in intrusions against organisations including the US aerospace sector, and a tool based on PentAGI used to scan organisations across Asia.

Centre · 1John Hultquist, chief analyst at Google's Threat Intelligence Group, said attackers are mostly not using groundbreaking methods but are still going after known, unfixed issues, "just with much greater speed and scale". He said security teams would need more automated defences to keep pace. Centre · 1The report also covered criminals stealing access to paid AI services and reselling it. Some victims whose cloud accounts were hijacked faced more than $100,000 in fraudulent charges in a single day, and cumulative costs reached $1 million in the worst cases. Centre · 1Google said North Korean-linked hackers and other groups were building AI into their operations, and that state-backed groups were using AI to create fake online identities and political content.

Centre · 1Google noted limits to the tools: they often leave traces defenders can spot, and some AI-generated exploits failed. It expects them to improve and is developing its own AI security tools to find and fix vulnerabilities first. Right · 1The New York Post linked Google's warning to outages that hit major US institutions in the past week, saying federal officials and experts suggested AI was to blame. Newsweek's account does not mention these outages, and the Post excerpt gives no detail on the cause.

Every sentence links to the reporting it rests on. The pill in front of each says where its sources sit: Left, Centre or Right when one side supplies at least half of them, Mixed when they are evenly split. The number is how many outlets it cites.

Left0 outlets

No left outlet in our sources has covered this story yet.

Centre1 outlet

Framing
A detailed explainer of Google's AI Threat Tracker report, with named tools, victims, costs and limits, and with measured quotes from Google's chief analyst.
Emphasis
Technical detail, financial losses to victims, Google's caveats about the tools' weaknesses and its own defensive AI work.
Leaves out or plays down
Does not say the actors were Russian or call this a historic first, and does not mention the recent US outages.
Charged language
“The era of autonomous attacks has arrived”
For example
“In a statement accompanying the report, Google said: "The era of autonomous attacks has arrived."” — Newsweek
“Google said autonomous hacking tools still had weaknesses.” — Newsweek

Right2 outlets

Framing
A short, dramatic lead on Russian AI bots pulling off a historic first, tied to current US outages.
Emphasis
Russian attribution, the 'first-ever' claim, and a link to outages hitting US institutions.
Leaves out or plays down
Leaves out the report's detail on other tools, French targets, fraud costs, Google's caveats and its defensive plans.
Charged language
“Russian AI bots”“first fully automated cyber attack”“institutions that run America”
For example
“Russian hackers put together an AI agent that appears to have pulled off the first-ever fully automated cyber attack in history, according to Google.” — New York Post
“Google’s warning comes as outages hit major institutions that run America this past week, with federal officials and experts suggesting AI is to blame.” — New York Post

Ask about this story

Loading…