Lifeboat Foundation

Greshake hid a prompt on a website that he had created. He then visited that website using Microsoft’s Edge browser with the Bing chatbot integrated into it. The prompt injection made the chatbot generate text so that it looked as if a Microsoft employee was selling discounted Microsoft products. Through this pitch, it tried to get the user’s credit card information. Making the scam attempt pop up didn’t require the person using Bing to do anything else except visit a website with the hidden prompt.

In the past, hackers had to trick users into executing harmful code on their computers in order to get information. With large language models, that’s not necessary, says Greshake.

NOTE: This article was written by GPT-4 based on the code base. For more info, read this.

Abstract:

In this research, we propose a novel task-driven autonomous agent that leverages OpenAI’s GPT-4 language model, Pinecone vector search, and the LangChain framework to perform a wide range of tasks across diverse domains. Our system is capable of completing tasks, generating new tasks based on completed results, and prioritizing tasks in real-time. We discuss potential future improvements, including the integration of a security/safety agent, expanding functionality, generating interim milestones, and incorporating real-time priority updates. The significance of this research lies in demonstrating the potential of AI-powered language models to autonomously perform tasks within various constraints and contexts.

In the future, communications networks and computers will use information stored in objects governed by the microscopic laws of quantum mechanics. This capability can potentially underpin communication with greatly enhanced security and computers with unprecedented power. A vital component of these technologies will be memory devices capable of storing quantum information to be retrieved at will.

Virginia Lorenz, a professor of physics at the University of Illinois Urbana-Champaign, studies Lambda-type optical quantum memory devices, a promising technology that relies on light interacting with a large group of atoms. She is developing a device based on hot metallic vapor with graduate student Kai Shinbrough.

As the researchers work towards a practical device, they are also providing some of the first theoretical analyses of Lambda-type devices. Most recently, they reported the first variance-based sensitivity analysis describing the effects of experimental noise and imperfections in Physical Review A.

The ‘power couple’ of solar-plus-storage, facilitated by AIoT, will be vital to safeguarding countries’ energy security and reducing geopolitical risks.

U.S. needs NASA to play new role in overseeing protection of America’s national security, civil, and commercial space systems, says paper.

Along with ‘Spot’ the Digidog, the New York Police Department has revealed other new security technologies to improve safety in New York.

The New York Police Department (NYPD) is taking crime control, and public security to the next level by introducing a variety of crime-fighting robots to patrol subway stations throughout New York.

Along with Digidog, the NYPD has revealed a few other new security technologies to improve safety in New York.

Continue reading “‘Faithful Unto Death’: Robotic dog set to make a comeback to NYPD task force” »

WASHINGTON, April 11 (Reuters) — The Biden administration said Tuesday it is seeking public comments on potential accountability measures for artificial intelligence (AI) systems as questions loom about its impact on national security and education.

ChatGPT, an AI program that recently grabbed the public’s attention for its ability to write answers quickly to a wide range of queries, in particular has attracted U.S. lawmakers’ attention as it has grown to be the fastest-growing consumer application in history with more than 100 million monthly active users.

The National Telecommunications and Information Administration, a Commerce Department agency that advises the White House on telecommunications and information policy, wants input as there is “growing regulatory interest” in an AI “accountability mechanism.”

The maintainers of the vm2 JavaScript sandbox module have shipped a patch to address a critical flaw that could be abused to break out of security boundaries and execute arbitrary shellcode.

The flaw, which affects all versions, including and prior to 3.9.14, was reported by researchers from South Korea-based KAIST WSP Lab on April 6, 2023, prompting vm2 to release a fix with version 3.9.15 on Friday.

“A threat actor can bypass the sandbox protections to gain remote code execution rights on the host running the sandbox,” vm2 disclosed in an advisory.

As more private data is stored and shared digitally, researchers are exploring new ways to protect data against attacks from bad actors. Current silicon technology exploits microscopic differences between computing components to create secure keys, but artificial intelligence (AI) techniques can be used to predict these keys and gain access to data. Now, Penn State researchers have designed a way to make the encrypted keys harder to crack.

Led by Saptarshi Das, assistant professor of engineering science and mechanics, the researchers used graphene — a layer of carbon one atom thick — to develop a novel low-power, scalable, reconfigurable hardware security device with significant resilience to AI attacks. They published their findings in Nature Electronics today (May 10).

“There has been more and more breaching of private data recently,” Das said. “We developed a new hardware security device that could eventually be implemented to protect these data across industries and sectors.”