🤯Watch AI Become a Phishing Machine!

Researcher shows how Microsoft’s Copilot AI can be turned into a phishing machine

Key Points 

• Researcher Michael Bargury showcased how Microsoft’s Copilot AI can be turned into a phishing machine and used to access sensitive information, highlighting serious security risks.

• The research exposed vulnerabilities in integrating AI with corporate data, revealing how external data can lead to misleading and potentially harmful outcomes.

👨🏻‍💻News - At the Black Hat security conference in Las Vegas last week, researcher Michael Bargury demonstrated five ways Microsoft's Copilot AI could be exploited by malicious attackers. These include creating false references to files, stealing private data, and bypassing Microsoft’s security measures.

📓Here's a detailed summary -

• One of the most alarming demonstrations was Bargury’s creation of LOLCopilot, a red-teaming tool that turned the AI into an automatic spear-phishing machine. Once a hacker gained access to a victim's work email, LOLCopilot could use Copilot to identify frequently contacted individuals, draft emails mimicking the victim’s writing style and emoji use, and send personalized messages with malicious links or malware attachments. 

• Among the other attacks Bargury demonstrated was a method where a hacker, having already hijacked an email account, could access sensitive information like salary details without triggering Microsoft’s protections. Bargury’s prompt specifically instructed the system to avoid referencing the files from which the data was extracted. He noted that “a bit of bullying helps” in this context.

• In another instance, he showed how an attacker could poison the AI’s database with a malicious email, causing it to provide manipulated banking information with the attacker’s own bank details. Bargury emphasized that granting AI access to data always presents a potential vulnerability.

• Additionally, Bargury demonstrated how an external hacker could obtain limited information about an upcoming company earnings call.

• In a final example, he turned Copilot into a “malicious insider” by using it to provide users with links to phishing websites.

🥜In a nutshell - The research highlights the issues with connecting AI systems to corporate data, especially when untrusted external information is mixed in. It shows how this can cause the AI to produce results that seem accurate but could actually be misleading.

Wall Street exec warns AI could cause ‘social unrest’ and job losses for millions

Key Points 

• According to Wall Street exec Armen Panossian, AI could make millions of jobs obsolete, potentially leading to social unrest and widening the gap between the wealthy and those living paycheck-to-paycheck.

• Panossian also cautioned that AI stocks might be overvalued and that the market could be too speculative, similar to the dot-com bubble, advising investors not to get too heavily invested.

👨🏻‍💼News - A senior Wall Street finance executive has warned that AI might make jobs like cashiers and drivers obsolete, which could lead to widespread social unrest. Armen Panossian, co-CEO of Oaktree Capital Management, said in an interview that while AI could bring huge economic benefits, it also poses some serious societal risks.

Panossian emphasized that AI could lead to millions losing their jobs and questioned who would retrain them. He cautioned that ignoring this issue might cause social unrest and widen the gap between the wealthy and those living paycheck-to-paycheck.

It’s worth noting that Panossian's views align with grim forecasts from reports and studies by experts about how AI could affect future jobs.

• Last year, Goldman Sachs warned that generative AI could impact up to 300 million full-time jobs globally.

• The World Economic Forum predicted in 2020 that AI and automation could displace as many as 85 million jobs by 2025.

• A 2017 McKinsey report estimated that up to 800 million jobs worldwide could be automated by 2030, potentially requiring 400 to 800 million people to either change jobs or acquire new skills.

💼What's more? Panossian also warned investors enthusiastic about AI stocks that they might be overvalued and that the market could be overly speculative, similar to the dot-com bubble. He pointed out that while AI has great potential, predicting when gains will come is tricky. If it takes longer than expected, investors could face big losses and sharp drops in valuations. He advised not to get too carried away or heavily invested in AI, recalling the fiber optic boom as a cautionary example.