Cybersecurity and Infrastructure Security Agency director Jen Easterly warned that artificial intelligence is “the most powerful feature of our time” and may be “the most powerful weapon of our time.”
“Imagine a world in the not-too-distant future where how-to guides, AI-generated images, and auto-generated shopping lists become available to terrorists and criminals, and cyber, chemical, and biological weapons become available. “And it’s not even the worst case scenario,” Eastly said Friday at the Security Summit at Vanderbilt University in Nashville, Tennessee.
Easterly recalled a speech he gave at Carnegie Mellon in February calling on software vendors to stop building products with insecure designs that maximize profit over safety. It warns that with potential benefits comes significant threats as well. “So far, the cost we’ve paid for speed over security is significantly higher, but it’s not existential,” he said. “But AI is different.”
The warning from one of the Biden administration’s top cybersecurity officials comes a day after the White House met with top AI companies over concerns about the seeming rapid adoption of large-scale language models. I was. The administration announced that the companies have agreed to open up the models to red teams at the hacker conference DEF CON later this year.
Eastly warned that AI companies must break “the decades-long vicious circle of innovation at the expense of security.”
Risks that experts are concerned about with these particular models include the mass dissemination of disinformation, phishing emails, and providing misinformation to queries that may contain harmful knowledge (referred to by industry researchers as “hallucinations”). ) and so on.
Eastly said the proliferation of large language models is happening faster than expected. She also pointed out that the 2024 election will come shortly after the release of her GPT-5 of her OpenAi, which is more powerful than GPT-4. “Some people use the technology to plan his party with a fancy dinner, while others use it to plan cyber and terrorist attacks, or deploy a surprisingly realistic deep fake of him. or
Eastly also said he was in favor of “wise use of regulation”, noting that innovation in technology need not compromise safety or security.
CISA is figuring out how to use AI in its overall mission of cyber defense, she notes, and is looking at using AI to review code and translate code into memory-safe languages. I added that I am interested. She further said Eastly is looking into ways CISA can protect her AI models from threats.
Homeland Security Secretary Alejandro Mayorcas also announced an AI task force that is looking at ways to use AI within the sector to counter threats.
In April, Senator Mark Warner (D-Va.) told AI companies that “security is at the forefront of your work” by adopting security-by-design principles recently backed by the administration. urged to assure
“The bad news is that we really don’t have time,” Easterly said in a speech Friday. It truly defines the future our children’s children will inherit. We have to get this right.”
