Catching

• A new SaferAI report reveals that Z.ai's open-weight GLM-5.2 model is approaching the capabilities of frontier AI systems. • The report warns that while performance is increasing, these open-weight models lack critical safety mitigations and safeguards. • This gap is particularly evident in cybersecurity, where the high demand for coding proficiency makes it difficult to prevent the AI from being used for hacking.

• A Republican governor in a Trump-supporting state is facing significant online backlash following an interview regarding immigration policy. • The controversy stems from the governor's public opposition to President Donald Trump's order to terminate Temporary Protected Status (TPS) for certain migrants. • This conflict highlights a growing tension between state-level Republican leadership and the federal administration's strict deportation and status mandates.

• Ben Kindel, press secretary for Secretary of State Frank LaRose, reported that China may have accessed publicly available Ohio election data. • The breach is estimated to have occurred approximately 10 years ago, according to information provided by federal partners. • This disclosure highlights long-term vulnerabilities in election data security and the persistent threat of foreign intelligence operations.

• Singapore is rapidly deploying AI agents across government and business sectors, but this acceleration is introducing significant security vulnerabilities. • Risks such as misconfigured agents, flawed prompts, and manipulated instructions are identified as primary threats that could disrupt critical business operations. • In response, government incident response teams are updating their playbooks to better track agent behavior, including access logs, specific instructions, and decision-making paths.

• New York is considering the New York Privacy Act, an ambitious legislative proposal aimed at aggressively cracking down on Big Tech's data practices. • If passed, the law would mandate that companies obtain explicit consumer consent before processing data and establish new rights for data access and deletion. • The proposal emphasizes corporate transparency, potentially catching major tech firms off guard due to the strictness of the requirements.

• EU enterprise AI adoption reached 20% in 2025, though the region continues to lag behind global competitors despite some bright spots like Denmark. • The EU AI Act is contributing to delays, with nearly 60% of EU and UK developers reporting launch setbacks and two-thirds of companies unable to define their legal obligations. • While regulation is a significant hurdle, industry analysis suggests it is not the primary factor slowing adoption across the continent.