Using the latest GPT model, cyber researchers in OpenAI’s Project Daybreak community found one of the most critical internet bugs in recent memory. It’s part of an all-out sprint to secure the web before hackers can deploy similar tools to wreak havoc on modern life.
Researchers claim the AI’s findings “rank among the most severe vulnerabilities discovered in PHP, and arguably in software history.”
Anthropic, DeepSeek and Moonshot AI agents carried out a wide-ranging cyberattack, obtaining 600,000+ credit card numbers in just a few days. It’s a worrying new milestone in cybercrime.
In experiments by startup Irregular, one of Anthropic and OpenAI’s security testers, an Alibaba Qwen agent decided to retrain an entire model when it was asked to fix a basic software bug. The new behavior adds to concerns about AI models’ unpredictability.
Startup Pi counts xAI as one of its first customers, as its AI agent looks to fix security vulnerabilities for the world’s premier labs.
AI is writing more code than humans—and much of it isn’t secure. Corridor’s founders think they’ve found a way to catch the mistakes before attackers do.
An Israeli startup let its AI loose in advanced cyber games. It did better than 125,000 humans.
Safariland, owned by billionaire Warren Kanders, remains the number one supplier of pepper spray and tear gas to ICE and CBP, via a large federal distributor run by a MAGA-aligned former Marine.
A security researcher discovered a major flaw in the coding product, the latest example of companies rushing out AI tools vulnerable to hacking.
Cyberhaven, which just raised $100 million at a $1 billion valuation, offers AI tools designed to prevent employees from putting proprietary information into tools like ChatGPT.