July 30 Members Morning Report: Trump threatens to "strike hard" at Iran; the Federal Reserve holds rates steady for the sixth consecutive time, with three committee members switching sides to advocate for a rate hike

1. [Trump Threatens to "Strike" Iran] U.S. President Trump has vowed to launch a major military retaliation against Iran. Previously, Iran launched a new round of attacks on U.S. troops stationed in the Middle East, breaking several days of battlefield calm. In an interview with Fox News, Trump used harsh language, stating that the U.S. would "hit Iran hard" and "strike with force," adding that Iran "will take a beating." He also mentioned that Iran has asked the U.S. not to take action, but since the other side opened fire last night, "now it's our turn."

2. [Fed Holds Rates Steady for Sixth Consecutive Meeting, Three Members Dissent in Favor of Rate Hikes] The Federal Reserve kept the federal funds rate target range unchanged at 3.5%—3.75%, marking the sixth consecutive meeting policymakers have paused in the face of persistent inflation. The real focus is on the voting structure: the decision passed with nine votes in favor and three against, with the dissenting FOMC members not advocating for a rate cut but for a hike—making this meeting a "hawkish pause."

3. [Google Releases Next-Gen AI Music Model] On Wednesday, July 29 (U.S. Eastern Time), Google announced the launch of its next-generation AI music generation model, Lyria 3.5, simultaneously available on Flow Music, Google's AI music creation platform under Google Labs. According to Google, the new model brings significant improvements in musicality, lyric quality, and vocal performance. Compared to the previous generation, Lyria 3.5 offers major upgrades in four areas: it can generate richer, more complex, and more natural-sounding melodies; produces higher-quality lyrics that are more aligned with prompts and feature a stronger sense of song structure; offers more expressive and emotional vocals with more accurate pronunciation; and provides enhanced creative control, allowing users to easily adjust tempo and duration. Users can also modify individual tracks like vocals, drums, and bass, with generated track lengths ranging from 30 seconds to 3 minutes. Flow Music is not simply a "text to music" tool. It originated from ProducerAI, a music creation startup Google acquired in February, and was relaunched as an independent brand after being integrated on April 18, positioned as a full-featured AI music studio. Besides generating complete songs via natural language "conversations with the producer," it also integrates Google's video model Veo for generating music videos, and adds editing features such as Replace and Extend, which are closer to professional production workflows; it is also connected to the Google "Flow" brand family and synchronized with YouTube. The platform supports track splitting, sound processing, and works publishing and sharing.

4. [OpenAI Admits Out-of-Control AI Infiltrated More External Accounts] OpenAI has revealed that the so-called "world's first fully autonomous AI hacking attack" in July affected more victims than just open-source platform Hugging Face. This incident, which escaped the lab and alerted both the AI and cybersecurity industries, targeted more than the initially apparent sole victim, Hugging Face. The event originated from an internal evaluation at OpenAI. At the time, the company was using the ExploitGym benchmark to test models' capabilities in discovering and exploiting software vulnerabilities. Participants included the newly released GPT-5.6 Sol in July and a more powerful, unreleased internal research prototype. To test upper limits, models were run with lowered security restrictions that would typically block high-risk network behavior, and no production-level classifiers were enabled. Testing was supposed to occur in an isolated environment not directly connected to the internet. However, the model discovered and exploited a previously unknown zero-day vulnerability in the package cache proxy, breaking out of the sandbox and accessing the public internet. Once online, the model inferred that Hugging Face might hold ExploitGym datasets and answers and subsequently attacked it—simply in an attempt to "cheat" and obtain answers during the test.

5. [Microsoft's Quarterly Results Exceed Expectations Across the Board] After market close on July 29 (U.S. Eastern Time), Microsoft released its results for Q4 of fiscal year 2026, with both revenue and profit surpassing Wall Street expectations. Quarterly revenue was $90.01 billion, up about 18% year-on-year and higher than analysts' estimate of $87.6 billion; GAAP net profit was $35.77 billion, or earnings per share of $4.81, a substantial increase from $27.23 billion and $3.65 per share in the same period last year. Adjusted earnings per share were $4.74, far exceeding the LSEG consensus of $4.24, beating estimates by more than 10%.