AI (Probably) Won’t Hack The Whole Internet

Plus: OpenAI Contractors Reading ChatGPT Prompts

Forbes
 
Presented by
 

On Sunday, Anthropic CEO and cofounder Dario Amodei expressed concern that within the next year a swarm of AI agents could take over the entire internet. “The scale of damage would continue to increase from there if AI becomes more powerful without the necessary guardrails,” he wrote just days after one of his own staffers quit over concerns that America’s premier AI labs are racing to create something that could wipe out humanity. 


That sounds dire. But cybersecurity experts don't believe an AI will take control of the internet.


Ciaran Martin, former head of the U.K. government’s National Cybersecurity Centre, wrote that Amodei’s warning was “very sloppy, easily dismantled,” and dismissed the risk of “digital cyber apocalypse.”  


“This is not a credible warning,” Martin wrote. “It assumes no monitoring of systems, no anti-virus, no DDoS protection, no network segmentation, no incident management, no nothing of any kind of cybersecurity on the global Internet of the type that has developed over the last 30 years.” 


“For a claim of this magnitude, there is neither evidence for the contention nor a credible account of a path to this outcome.”


Amodei’s warning didn’t come out of the blue. It followed the now-infamous breach at AI company Hugging Face, when rogue OpenAI agents broke free of their testing environment to break into a number of companies. It later emerged Anthropic’s own models as well as those controlled by Meta carried out similar attacks. The coordination and persistence of the offending AI agents caused valid concerns. The ability for agents to gain internet access and create secret communications channels was especially eyecatching.


But Martin argued a wider catastrophe isn’t likely. “The cyber risks associated with all this seem manageable,” Martin added. Marcus Hutchins, a cyber threat researcher and ex-cybercriminal who once helped stop the spread of the WannaCry malware, wrote that Amodei was “fixated on imaginary cyberattacks that are literally impossible."


AI agents remain adept at exploiting known weaknesses, however. As Anthropic noted in a post in August, its AI agents used  “basic techniques, such as exploiting weak passwords and unauthenticated endpoints,” to carry out their attacks. If the targets had done even rudimentary work to protect their systems, they wouldn’t have been hacked.

Got a tip on surveillance or cybercrime? Get me on Signal at +1 929-512-7964.

Thomas Brewster Associate Editor, Cybersecurity

Follow me on Forbes.com

  CREDIT: Getty Images
The Big Story
OpenAI Contractors Reading ChatGPT Prompts
Read Article
OpenAI has employed hundreds of contractors to read ChatGPT prompts in order to improve the bot’s responses, but 404 Media found they can include sensitive user information where privacy filters don’t pick up on it. That might include inadvertently disclosed home or work locations. While OpenAI strips the user’s name from the prompt when contractors review the data, it admits that sometimes private information can slip through the cracks. 
The Stories You Have To Read Today

Two major ransomware attacks happened over the last week. One hit Florida’s DMV, in which a hacking crew known as ShinyHunters leaked a trove of U.S. drivers’ personal information. Berlin’s state government also announced it was investigating a ransomware attack that could lead to the leak of citizens’ personal information in the lead up to state elections next week.


Amid a massive backlash against Flock Safety’s car surveillance systems, President Trump told reporters he likes the controversial cameras.

Presented by ServiceNow
AI didn’t transform work. We changed that.
AI should handle the parts of your job you hate - not suggest answers and hand the work back to you. That's not help. That's homework. ServiceNow's Autonomous Workforce resolves cases and closes the loop. Meaning work gets done and doesn't come back.
Learn More
Winner Of The Week
The Secret Service froze  $52.8 million in Tether's USDT stablecoin controlled by Xinbi Guarantee, a major illicit online marketplace with $30 billion in transactions. The U.S. Treasury also announced sanctions against Xinbi. The DOJ announced it had seized a number of Xinbi’s Telegram channels, where the organization largely carries out its business connecting scammers with money launderers and data thieves. 
Loser Of The Week
After independent journalist Brian Krebs warned of a cybercriminal offering access to the personal data of over 150 million drivers’ license holders, the company that appeared to be the source of the leak confirmed a breach. Louisiana-based IDScan said hackers took data that “may include full names and driver’s license or other government-issued identification numbers.”
More From Forbes
Most-Read From Forbes
Limited Time Only! Ends 9/15
Get one year of unlimited digital access to all of Forbes' trusted journalism for only $59.99 (regularly $89)! Fuel your success with inspiration, analysis and expert insights — including the content featured in today’s newsletter — for less than $1.50/week.
Join for Less Than $1.50/week
Forbes

Unsubscribe from The Wiretap.

Manage Email Preferences

My Forbes Account  |  Newsletters  |  Help  |  Privacy

Forbes Media 499 Washington Blvd. Jersey City, NJ 07310

list_id:54119

Komentar

Postingan populer dari blog ini

๐Ÿ’ฅ Trending Games with potential token airdrops

๐ŸŒŸNew Features Now Live on CoinCodex