Recent AI models have demonstrated impressive mathematical capabilities, discovering counterexamples to long-standing conjectures and finding novel applications of existing techniques—achievements that fall into relatively constrained categories of problem-solving. However, current AI systems remain limited in their ability to develop entirely new conceptual frameworks or deep theoretical structures, suggesting that while AI will likely continue advancing in mathematics, it is not yet approaching human-level mathematical creativity in the most demanding sense.
New research shows that country-of-origin labels can obscure an AI model’s upstream dependencies, inherited behaviors and potential security risks. The post Think You’ve Eliminated Chinese AI? Check the Model’s Lineage, Cisco Says appeared first on SecurityWeek.
Here’s a look at the most interesting products from the past week, featuring releases from A10 Networks, Abnormal AI, F5 Networks, Intezer, Netscout, ScienceLogic, Searchlight Cyber, SelectHub, ServiceNow, Snyk, Tanium, and Tufin. ServiceNow organizes autonomous security around six solution areas ServiceNow has announced an acceleration of its Autonomous Security vision with six unified solutions that help deliver prevention-first, AI-native cyber defense across unified exposure management, continuous vulnerability detection, cyber-physical security, identity and access security, and … More → The post New infosec products of the month: August 2026 appeared first on Help Net Security.