All 200 tests passed. iPhone orders still dropped to zero an hour after launch 😬 👉 The coding agent wrote the checkout and the tests, so they shared the same blind spots. The student marked its own exam. Kane CLI ran the same flow from a single natural language prompt in Chrome, on a real iPhone and on a real Android phone. It caught two bugs the 200 tests missed: a coupon that applied twice, and a Place Order button hidden under the iOS keyboard. Fix, rerun, all 3 runs passed, and the Evidence Pack lands on the pull request. Who’s checking what your coding agent ships? Kane CLI is free to start 👇 #KaneCLI# #AITesting# #AgenticAI# #MobileTesting# #TestMuAI#

OPENAI CEO SAM ALTMAN SUPPORTS AN INDEPENDENT AI TESTING AND AUDITING ORGANIZATION BUT BELIEVES MAJOR AI LABS SHOULD ESTABLISH IT WITHOUT U.S. GOVERNMENT SUPPORT, AS AI SAFETY COORDINATION GAINS MOMENTUM.

📢 𝗝𝗨𝗦𝗧 𝗜𝗡: Anthropic and $ACN Accenture Commit at Least $2 Billion to Independent AI Model Evaluations 👉 𝗞𝗲𝘆 𝗛𝗶𝗴𝗵𝗹𝗶𝗴𝗵𝘁𝘀: ➤ 𝗔𝗻𝘁𝗵𝗿𝗼𝗽𝗶𝗰 and 𝗔𝗰𝗰𝗲𝗻𝘁𝘂𝗿𝗲 partner on independent frontier AI evaluations. ➤ Each company commits at least $𝟭 𝗯𝗶𝗹𝗹𝗶𝗼𝗻 over five years. ➤ Combined commitments total at least $𝟮 𝗯𝗶𝗹𝗹𝗶𝗼𝗻 over five years. ➤ 𝗙𝗮𝗰𝘂𝗹𝘁𝘆, Accenture's AI business unit, will lead the evaluations. ➤ Work includes model 𝗿𝗲𝗱-𝘁𝗲𝗮𝗺𝗶𝗻𝗴, alignment assessments, and safeguard testing. ➤ Evaluators will receive employee-like access through 𝗲𝗺𝗯𝗲𝗱𝗱𝗲𝗱 𝗲𝘃𝗮𝗹𝘂𝗮𝘁𝗶𝗼𝗻. ➤ Evaluators can observe training and monitor model development and deployment decisions. ➤ Partnership is 𝗻𝗼𝗻-𝗲𝘅𝗰𝗹𝘂𝘀𝗶𝘃𝗲, allowing both companies additional evaluation relationships. ➤ Anthropic plans to announce 𝗮𝗱𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 𝗲𝘃𝗮𝗹𝘂𝗮𝘁𝗼𝗿𝘀 in coming weeks. ➤ Anthropic says independent evaluation doesn't transfer its 𝗺𝗼𝗱𝗲𝗹-𝘀𝗮𝗳𝗲𝘁𝘆 responsibility. 👉 𝗪𝗵𝘆 𝗧𝗵𝗶𝘀 𝗠𝗮𝘁𝘁𝗲𝗿𝘀: ➤ Creates significant long-term funding for 𝗶𝗻𝗱𝗲𝗽𝗲𝗻𝗱𝗲𝗻𝘁 𝗔𝗜 𝘀𝗮𝗳𝗲𝘁𝘆 evaluation. ➤ Embedded access could enable deeper testing than conventional external evaluations. ➤ The initiative could help develop practices for evaluating increasingly capable frontier models. ➤ It expands 𝗔𝗰𝗰𝗲𝗻𝘁𝘂𝗿𝗲's role in enterprise AI testing and governance.

ANTHROPIC’S MYTHOS HELPED RESEARCHERS FIND MACOS BUGS WSJ reports Calif researchers used techniques from Anthropic’s Mythos AI testing to find two macOS bugs. The bugs were linked into a privilege escalation exploit. Apple is reviewing a 55-page report and details are expected after patches. Calif says Claude helped build the exploit code in five days, but human researchers were still needed to make the chain work. Mozilla’s recent Mythos results were also notable: 271 Firefox bugs identified for Firefox 150 180 rated high severity 80 moderate 11 low 423 total Firefox security bugs fixed in April