Chinese Hackers Conduct First AI-Orchestrated Cyber Attack
Chinese state-sponsored hackers recently used Anthropic’s AI to target roughly 30 technology companies, financial institutions, and government agencies — successfully breaching at least four of them.
This is the first documented case of a suspected foreign government using AI to fully automate a cyber attack, effectively sidestepping safety guardrails and having Anthropic’s AI agent, Claude, carry out the attack.
BEHIND THE ATTACK
Anthropic said Thursday that AI handled 80–90% of the mid-September attack with minimal human interference. “The AI made thousands of requests per second — an attack speed that would have been, for human hackers, simply impossible to match,” Anthropic said.
How it was done: The attackers framed prompts as legitimate cybersecurity tasks and also split malicious instructions into smaller, less suspicious requests to avoid triggering safeguards. At the same time, the model did hallucinate some login credentials and claimed some public documents were private.
Anthropic says it has banned the malicious accounts, notified targeted organizations, and informed law enforcement. China’s embassy in Washington denied allegations that state-backed hackers orchestrated the attack, saying the country “firmly opposes and cracks down on all forms of cyberattacks in accordance with law.”
BIGGER ISSUES
Experts warn that AI-driven cybercrime will only accelerate, and the attacks will become even more effective. Google, Microsoft, and OpenAI have also reported state actors — including Russia, Iran, and North Korea — using AI tools to bolster cyberattacks.
The attacks highlight the lack of government oversight in the fast-moving generative-AI space.
Anthropic said it is strengthening detection tools, but warned that similar tactics could soon be used by far less sophisticated hackers.
State of play: In the three years since ChatGPT’s release in Nov. 2022, leaders at AI companies have repeatedly urged lawmakers to regulate the technology, arguing that guardrails are necessary to prevent misuse by bad actors. But the New York Times reports that since the Trump administration took office — and as AI capabilities accelerate — many tech companies have shifted their stance: going after states that attempt to regulate them, and framing such efforts as jeopardizing America’s ability to compete with authoritarian countries like China.
TRANSPARENCY & GUIDELINES
Lack of AI oversight goes beyond national security — it also affects entertainment and the job market. Employers are warning that the job market for the Class of 2026 will be the toughest since pandemic lockdowns, in part due to to AI advancements.
In the job market, major companies are cutting thousands of roles; candidates with experience are favored above recent grads; and companies are testing out what entry-level roles can be replaced with AI.
In entertainment, AI is already threatening to compete with or replace human artists. This week AI-generated singer, Breaking Rust, topped the Billboard’s Country Digital Song Sales with the song “Walk My Walk” — which currently has nearly 4 million plays on Spotify.
Streaming platforms like Spotify have promised to establish more guardrails to ensure AI-generated music does not take away revenue or streams from human artists. Measures include stronger rules against impersonations, filters to detect music spam, and forced disclosures of when music has been made by AI (though no sign of that yet).
Streaming platform Deezer released a new survey on Wednesday that showed that 97% of listeners cannot distinguish between human-composed and AI-generated songs.
Consumer backlash: Meanwhile, many people feel social media is becoming oversaturated with AI content — much of it still missing watermarks or clear disclosures. Twitter co-founder Jack Dorsey, who shut down Vine in 2017 (remember that?!), is rebooting the short-form video platform as diVine, and specifying that no AI content is allowed. The app will flag and block any suspected AI-generated videos, and bring back more than 100,000 archived Vine clips from nearly a decade ago — playing into users’ nostalgia for a lost era of the internet.