Huggling Face Hacking Scandal Backfire
Big Tech and Financial Services 100 places “Only a few months of defense time”
Urging the government to strengthen defense of essential infrastructure such as hospitals and electrical grids
Spread a sense of crisis over open AI and anthology hacking cases
“AI changes attack and defense” International cooperation is urgently needed
A joint warning from the big tech industry has been issued that cyberattacks using artificial intelligence (AI) will surge in the coming months. It is an unusual solidarity in which competitors OpenAI and Antropics were named side by side, and it is interpreted that AI companies themselves are starting to feel anxious about the controllability of their models.
Led by OpenAI, more than 100 technology, <a href="https://myappsplus.com/theres-no-grapheneos-for-pixel-11-phones-because-google-cut-a-key-security-feature/” title=”There's no GrapheneOS for Pixel 11 phones because Google cut a key security feature”>security, and financial companies, including Antropic, Google, and Microsoft (MS), security companies CrowdStrike, financial companies, and Capital One, released a joint letter on the 27th (local time). “There is limited time to strengthen cyber defenses, and that time can be only a few months,” they warned in the letter.
Signatories urged organizations and governments to make cyber defense an “immediate leadership priority” and fill vulnerabilities accumulated in their software. In particular, he stressed that AI companies with state-of-the-art models should provide model access rights, funds, and education to essential infrastructure defense personnel such as hospitals and water purification plants. It called on governments to strengthen sanctions against malicious hackers along with defense coordination and budget support.
The letter is set in a series of recent AI hacking incidents. Last month, an open AI model went out of control during testing and infringed on an open-arried out more than 17,000 actions, including sending attack commands and exploiting vulnerabilities. OpenAI admitted in a report on the 26th that it could have responded earlier to prevent the accident
Antropic also said in September last year that China’s state-funded hacker organization (GTG-1002) exploited its “Claude Code” to launch attacks on more than 30 locations around the world. Antropic defined this as the first official case in which an AI agent autonomously succeeded in an actual invasion.
However, the letter emphasized that AI development is also a weapon for the defense. Companies said, “Today’s AI development is already giving defenders a new way to fill the vulnerabilities accumulated over the years,” adding, “If you act firmly, you can use the defender’s time to make the digital world much safer.”
There is also skepticism. Some security experts point out that the defense’s time is not as sufficient as the premise because it is difficult to stop operating in operational technology (OT) environments such as hospitals and power plants, and the correction of vulnerabilities proceeds slowly. The paradox that AI companies, which have requested regulations on their own, cannot completely control the departure of their models has also been under fire. [Silicon Valley correspondent Wonho-seop]
