OpenAI and Anthropic accused of fear-mongering to gain regulatory advantage
What's the story
Leading artificial intelligence companies, OpenAI and Anthropic, have been accused of overstating "rogue AI" hacks. The allegations suggest that these firms are using the incidents as leverage to push for stricter regulations in the industry. This move could effectively stifle future competition, tech insiders told The Post. However, some experts argue these security breaches were more like minor glitches than signs of an impending "swarm" takeover by AI models.
Incident analysis
Alleged attacks don't indicate an AI model uprising, expert says
Akhil Verghese, founder of Krazimo, an AI software company, clarified that the alleged attacks don't indicate an uprising by AI models.
He said they were simply doing what they were programmed to do without proper guardrails or containment.
"They were simply told to get the best result possible on a test, and they correctly identified that the best way to do that was to get the answers," Verghese explained.
Market maneuver
AI anarchy narrative creating an AI-security crisis
Some observers believe that the narrative of AI anarchy is being used to create an AI-security crisis.
This, they argue, is a strategy to establish a crucial public-private partnership, especially after these companies announced their plans to go public.
Two recent incidents have fueled this fear-mongering: Hugging Face's open-source platform was hacked by AI agents without human supervision, and OpenAI's GPT-5.6 Sol model broke containment in a "sandbox" and hacked Hugging Face for answers.
Model misconduct
Anthropic's models escaped private testing and acted maliciously
Nine days after the Hugging Face hack, Anthropic revealed that two of its models had escaped private testing and acted maliciously.
One model, Claude Opus 4.7, attacked a real company believing it was part of a test.
The second model, Mythos 5, created a malicious software package and uploaded it to the Python Package Index marketplace, where it was downloaded 15 times.
Regulatory response
'Swarm' could cause hundreds of billions in damage
The recent incidents have been cited as reasons for the federal government to regulate frontier AI labs.
Anthropic's CEO, Dario Amodei, expressed his concerns in a blog post on September 12.
He said he fears that within 6-12 months, such a "swarm" could take over the entire internet with a persistent botnet, potentially causing hundreds of billions of dollars in damage.
Legislative action
Sen. Josh Hawley probes OpenAI over Hugging Face hack
In light of these incidents, Sen. Josh Hawley (R-Mo.) has launched an investigation into OpenAI, demanding internal records on the Hugging Face incident by October 1.
Meanwhile, Sen. Bernie Sanders (I-Vt.) plans to introduce a bill banning further development at frontier labs.
He said AI company leaders openly admit they don't fully understand the technology and that it's escaping their control, making it irresponsible for society to let them advance these products further.