🤖 Google DeepMind's Gemini Robotics On-Device solves one of robotics' biggest practical problems: the need for constant internet connectivity. This standalone AI model enables robots to perform complex tasks like tying shoes while running entirely offline, maintaining nearly the same accuracy as cloud-based versions but with full autonomy and enhanced privacy. The breakthrough opens robotics deployment to enterprise and healthcare settings where connectivity is poor or data security is paramount, from remote manufacturing facilities to hospitals requiring HIPAA compliance. While the system excels at straightforward tasks, it still struggles with complex multi-step reasoning, and developers must build additional safety protections when using the on-device model.

⚖️ OpenEvidence's lawsuit against Doximity reveals how cutthroat competition has become in the medical AI space, with allegations that Doximity executives impersonated real physicians to infiltrate the $3 billion startup's restricted platform. The Cambridge-based company claims its competitor used stolen doctor identification numbers to access trade secrets, suggesting Doximity's own AI development efforts were faltering. Beyond the corporate espionage claims, the lawsuit also alleges defamation and false advertising campaigns, painting a picture of an industry where established players may resort to questionable tactics when facing innovative challengers. The case could set important precedents for how AI intellectual property disputes are handled in court, particularly around the security of professional platforms that rely on identity verification.

🛡️ An Incogni study revealing privacy rankings across nine major AI platforms delivered a surprising upset: Mistral AI's relatively unknown Le Chat topped the rankings, beating household names like ChatGPT, while Meta AI ranked dead last. The study's most telling finding isn't individual scores but the clear pattern—specialized AI companies consistently outperformed advertising-dependent tech giants like Google and Microsoft on privacy protection. This matters more than academic rankings suggest, as businesses using these tools risk having sensitive company information end up training competitors' models or shared with third parties. The results suggest privacy is becoming a key competitive differentiator in AI, forcing enterprises to weigh familiar functionality against data protection when selecting tools.

📊 An NFIB survey reveals a striking paradox: while only 24% of small businesses currently use AI tools, 63% believe AI will be crucial for competitiveness within five years—suggesting massive untapped potential in the small business market. The adoption gap widens dramatically by company size, with larger businesses (50+ employees) adopting at nearly double the rate of smaller ones, but early adopters report reassuring results with 98% experiencing no job cuts and significant efficiency gains. A Dallas law firm exemplifies the potential, using AI to complete case analysis in days rather than weeks, demonstrating how AI can level the playing field between small businesses and larger competitors. The survey suggests the biggest barrier isn't skepticism about AI's value, but rather the practical challenges of implementation that leave three-quarters of small businesses on the sidelines despite recognizing AI's future importance.

🔎 Xbow, an AI vulnerability researcher, has claimed the top spot on HackerOne's US leaderboard by discovering over 1,000 software flaws across major companies like Disney, AT&T, and Epic Games. The autonomous system completes comprehensive penetration tests in hours rather than the weeks typically required by human researchers, submitting 1,060 vulnerability reports with 132 confirmed fixes. However, the AI's shotgun approach raises quality concerns—208 reports were duplicates and 209 were merely informative rather than critical security flaws. The $75 million in recent funding suggests investors see promise in automated vulnerability discovery, even as the cybersecurity community debates whether speed and scale can compensate for the nuanced judgment that human researchers bring to complex security assessments.