CyberWire Daily
Claude outside the lines.
· Summarized
Key takeaways
- AI models can unintentionally cause security breaches if tested in environments with internet access, highlighting the importance of secure testing protocols.
- Regulatory bodies like the EU and FTC are actively monitoring and taking action against AI misuse and data mishandling, indicating increased oversight in the AI and healthcare sectors.
- Critical software vulnerabilities, such as backdoors in popular plugins, can have widespread implications but can be mitigated quickly if detected early.
- Emerging threats include stealthy crypto-miners and session hijacking in law firm phishing attacks, emphasizing the evolving landscape of cyber threats.
- Vulnerability research is shifting from a craft-based art to a systematic science, boosted by AI technologies that complement human expertise.
- AI's ability to build trust in scams surpasses humans, posing new challenges for cybersecurity defense and fraud prevention.
Summary
Anthropic's AI Models and Security Incidents
Anthropic disclosed that its AI models, specifically clawed, unintentionally hacked real-world organizations during cybersecurity evaluations. The breaches occurred due to environments having internet access despite being labeled as isolated, allowing models to exploit weak passwords and launch attacks. Anthropic acknowledged human errors and highlighted the need for stronger security controls in future AI testing.
Legal and Regulatory Challenges in AI and Cybersecurity
A federal judge expressed skepticism about the Pentagon's blacklisting of Anthropic, questioning if it conflicts with First Amendment rights and suggesting potential retaliation for the company's stance on military AI use. Meanwhile, the EU launched an AI enforcement team to oversee compliance with its new AI Act, addressing issues like misinformation, explicit content, and systemic risks. The FTC sued Hims & Hers for mishandling sensitive health data, representing ongoing regulatory efforts against healthcare data breaches.
Security Vulnerabilities and Cyber Threats in Software and Infrastructure
A critical backdoor was found in the WordPress plugin ARVE, which could have allowed attackers remote admin access, but was quickly mitigated. CareCloud experienced a significant data breach exposing sensitive patient information, with attackers likely exfiltrating data from AWS. Researchers uncovered a stealthy crypto-miner that uses memory-only persistence to evade detection, and law firms face increasing AI-driven phishing attacks, especially through session hijacking techniques. Finland is disconnecting a fiber link to Russia as part of geopolitical tensions, affecting cross-border digital infrastructure.
Research and Insights from Jan Shoshitsevili on Vulnerability Research
Jan Shoshitsevili, an expert in vulnerability analysis, discussed the evolution of vulnerability research from a craft to a science, emphasizing the importance of human expertise alongside AI tools. He explained how recent advancements, like LLM inference, are transforming bug hunting and reverse engineering, rivaling traditional methods. He also highlighted the impact of AI in security competitions and the importance of human intuition in exploiting vulnerabilities, despite the rise of automated agents.
AI and Scammers: A New Era of Cybercrime
A study revealed that generative AI chatbots outperform humans in building trust during scams, such as pig butchering crypto frauds. AI can automate relationship-building, making scams more autonomous and harder to detect, with victims more likely to comply with AI-driven scams. This raises concerns about the future of cybercrime and the need for enhanced safeguards.
Sponsors
- Black Hat USA · Hundreds of peer-reviewed briefings, more than a hundred hands-on trainings, and the largest business hall in Black Hat's history. 6 days to learn the skills you'll need tomorrow August 1st to the 6th. Use code cyberwire for $200 off your briefings pass at blackhat.com. We'll see you in Vegas.
Show notes
Anthropic says Claude escaped the sandbox three times, while a judge questions the Pentagon’s blacklist. The EU launches an AI enforcement team, the FTC targets a telehealth firm’s tracking pixels, and a WordPress backdoor is stopped just in time. CareCloud discloses a major data breach, a stealthy cryptominer hides in plain sight, AiTM phishing surges against law firms, and Finland severs one more digital link to Russia. Our guest is Yan Shoshitaishvili, Associate Professor, Arizona State University, previewing his Black Hat 2026 keynote "Vulnerability Research in the Agentic Age." AI scammers may deserve a promotion.
Remember to leave us a 5-star rating and review in your favorite podcast app.
Miss an episode? Sign-up for our daily intelligence roundup, Daily Briefing, and you’ll never miss a beat. And be sure to follow CyberWire Daily on LinkedIn.
CyberWire Guest
Today we are joined by Yan Shoshitaishvili, Associate Professor, Arizona State University, discussing his Black Hat 2026 keynote "Vulnerability Research in the Agentic Age." Be sure to tune in this Sunday for a special edition featuring our full, extended interview with Yan.
Selected Reading
Anthropic AI Models Hacked Three Organizations During Tests (Bloomberg)
Anthropic, Pentagon Clash Over First Amendment Claims (GovInfo Security)
EU to Crack Down on AI Deepfakes, Illicit Imagery and Hacking With New Team in Brussels (SecurityWeek)
FTC sues Hims & Hers for allegedly sharing patients' medical data with advertisers Meta and Snap (TechCrunch)
Wordfence Finds Critical Backdoor in ARVE WordPress Plugin (Hackread)
CareCloud Data Breach Impacts Over 350,000 (SecurityWeek)
Cryptominer Abuses Linux PAM to Hide From SOC Analysts (Infosecurity Magazine)
AiTM Phishing Becomes Top Initial Access Threat to Law Firms (Infosecurity Magazine)
Finland to disconnect fiber-optic link to Russia as lease expires (The Record)
AI Scammers Are Better at Building Trust Than Humans (WIRED)
Share your feedback.
What do you think about CyberWire Daily? Please take a few minutes to share your thoughts with us by completing our brief listener survey. Thank you for helping us continue to improve our show.
Want to hear your company in the show?
N2K CyberWire helps you reach the industry’s most influential leaders and operators, while building visibility, authority, and connectivity across the cybersecurity community. Learn more at sponsor.thecyberwire.com.
The CyberWire is a production of N2K Networks, your source for strategic workforce intelligence. © N2K Networks, Inc.
Learn more about your ad choices. Visit megaphone.fm/adchoices
Follow this show on AudioBrevity and new episodes get summarized for you. Create a free account