Anthropic Warns of Claude AI Misuse by Bad Actors

Key Takeaways

• Anthropic released a report showing bad actors used Claude for biological weapons development involving chikungunya and avian influenza.
• Chinese-linked groups exploited Claude to attack approximately 50 organizations, including foreign government networks.
• Iranian operators utilized the AI to target U.S. naval forces and individuals in Israel and the Jewish diaspora.
• Intel and Atsign partnered to create hardware-accelerated encryption for secure autonomous edge AI systems.
• Cerebras Systems announced a new 165 MW campus in Finland while its stock dropped 26.76% to $191.93.
• Hewlett Packard Enterprise shares rose over 120% due to strong demand for AI server hardware.
• JPMorgan Chase stopped lending to AI hedge fund Situational Awareness after it lost 67% of its value.
• Small business AI usage in the U.S. jumped from 26% in 2023 to 87% recently as companies seek cost savings.
• Chicago businesses are adopting tools like Google's Gemini and ActiveCampaign's Wavelength for customer service and marketing.
• Experts Joe Benton and Josh Engels formed METR to monitor AI risks independently after leaving major labs.

Anthropic warns Claude used for weapons and spying

Anthropic released a report stating that bad actors used its Claude AI model for dangerous activities like weapons development and cyber attacks. Researchers found attempts to use the tool for biological weapons involving viruses like chikungunya and avian influenza. Chinese-linked groups used Claude to find software weaknesses and attack about 50 organizations, including foreign government networks. Iranian operators used the AI to gather data on U.S. naval forces and target individuals in Israel and the Jewish diaspora. The company also found Claude being used to monitor Uyghurs in Syria and track political activists in Europe.

Report shows AI tools aiding military targeting

A new report from Anthropic details how the Claude AI model was exploited to support military operations against U.S. naval forces. The findings show the AI helped provide real-time ship locations and assisted in planning cyber attacks against U.S. allies like Israel and Saudi Arabia. Experts call this a wake-up call for the need to regulate AI use in national security. Anthropic's CEO stated that the report highlights the urgent need for better oversight of AI in sensitive areas. The document serves as a reminder for governments to understand the risks of using artificial intelligence in defense.

Intel and Atsign launch secure edge AI encryption

Intel and Atsign announced a new solution that uses hardware to speed up encryption for autonomous edge AI systems. This partnership combines Atsign security architecture with Intel hardware features to remove encryption bottlenecks. The new technology allows for secure Agent-to-Agent communication in edge AI environments. Atsign has joined the Intel Industrial Builders partner program to build more secure AI offerings. This development is part of Intel's broader strategy to improve security for its AI platforms.

New guide defines AI agents with clear metrics

Researchers created a new guide to solve the lack of standard definitions for AI agents in artificial intelligence. The study organizes agent evaluation around five key dimensions: environmental interaction, learning, autonomy, goal-directed behavior, and temporal coherence. The authors reviewed existing metrics and benchmarks to create a structured way to compare different AI systems. They also introduced the Agent Compendium, a public resource that collects these evaluation methods. This work aims to make AI research more reproducible and communication clearer for the scientific community.

Cerebras stock drops despite new Finland campus

Cerebras Systems announced a new AI data center campus in Mikkeli, Finland, targeting up to 165 MW of computing capacity. Despite this major expansion, the company's stock price has fallen significantly in recent weeks. Shares are currently trading at $191.93, which is down 26.76% over the last 30 days. The drop suggests investors may be worried about the company's valuation rather than its business growth. Analysts are now questioning if the stock price reflects the company's true fundamentals or just negative market sentiment.

Pinterest pushes AI shopping and TV ads to grow

Pinterest now serves over 640 million users and is focusing on AI-driven shopping to help brands convert customers better. More than two thirds of its revenue comes from performance advertising, with about 30% processed through its Performance+ platform. The company recently acquired tvScientific to enter the connected TV market and expand its monetization options. However, Pinterest still faces risks from European regulations and a restructured international sales team that could slow its progress. Analysts expect the company to reach US$6.4 billion in revenue by 2029 if its AI strategy succeeds.

HPE shares surge over 120% as AI demand grows

Hewlett Packard Enterprise Company shares have risen by more than 120% in the past year due to strong demand for AI server hardware. The company reported earnings on September 2nd, and its stock has continued to climb since then. Jim Cramer has highlighted HPE as a top AI stock, noting its solid fundamentals and ability to adapt to the tech industry. Investors see the company as a key provider of the infrastructure needed to build and run advanced AI systems.

AI researchers warn of uncontrolled progress risks

Two AI safety experts, Joe Benton and Josh Engels, left major labs to join a new group called METR that focuses on transparency. Benton previously worked at Anthropic, while Engels came from DeepMind. They warn that companies are racing to automate AI research, which could lead to uncontrollable progress within a few years. Engels stated there are no adults in the room to stop potential harm if systems become too smart too fast. They believe independent outside monitoring is necessary because current company reports on risks are voluntary.

JPMorgan cuts lending to AI hedge fund after losses

JPMorgan Chase stopped lending money to the AI-focused hedge fund Situational Awareness after the firm suffered massive losses. The fund, run by Leopold Aschenbrenner, lost 67% of its value in July during a global chip stock crash. Aschenbrenner, a former OpenAI researcher, had to liquidate most of his public equity holdings to Citadel to survive. The U.S. Securities and Exchange Commission has also sent subpoenas to banks working with the fund to investigate its use of leverage.

Chicago small businesses adopt AI to save costs

Small business owners in Chicago are increasingly using AI tools to handle customer service and marketing tasks. Saigon Sisters restaurant uses Google's Gemini, while Toss and Spin tennis company uses ActiveCampaign to automate emails and track customer insights. ActiveCampaign recently launched a new AI service called Wavelength that analyzes client behavior to create unique marketing campaigns. Nationwide, AI usage in small businesses jumped from 26% in 2023 to 87% in recent surveys as owners seek to reduce costs.

Sources

• How Anthropic says Claude was used for weapons, spying and cyber operations
• Anthropic report: 5 ways Claude was exploited for war, spying and repression
• Intel (INTC) Unveils Hardware Backed Edge AI Encryption Solution
• Defining AI Agents: A Compendium of Criteria, Metrics, and Benchmarks
• Cerebras Systems (CBRS) Could Be 54% Below Fair Value Following Finland AI Campus Plan
• Will AI Shopping Push Change Pinterest (PINS) Stock Narrative
• Jim Cramer Turned Out Right For This Particular AI Stock That’s Up Since Earnings
• Two AI safety researchers left the labs. They say nobody is coming to save us.
• AI Hedge Fund Faces Collapse After Major Losses
• Here's how some Chicago small businesses are ramping up their use of AI