Reading

Leaders of three Republican-led committees in the U.S. House of Representatives have sent a letter to Nvidia CEO Jensen Huang, requesting his attendance at a joint closed-door briefing on November 18. The move follows recent reports alleging a close relationship between Huang and an executive identified as a Chinese intelligence asset.

According to CNBC, the letter was signed by House Foreign Affairs Committee Chairman Brian Mast, Select Committee on China Chairman John Moolenaar, and House Permanent Select Committee on Intelligence Chairman Rick Crawford. The lawmakers seek information on Huang's interactions with Alice Huang, an executive at Megaspeed, as well as Nvidia's security protocols against foreign adversary espionage. The New York Times previously reported that U.S. intelligence agencies found Alice Huang had developed a close relationship with Jensen Huang while funneling information to Beijing.

An Nvidia spokesperson stated that regional cloud customers receive ordinary business information, not sensitive trade secrets or national security-relevant data, and emphasized that meetings with customers are part of a CEO's job. The requested briefing date falls immediately after Nvidia's scheduled third-quarter earnings release on November 17, highlighting the company's exposure to regulatory scrutiny amid geopolitical tensions over AI chip exports.

Google (owned by Alphabet), Meta, and TikTok have filed separate lawsuits in US federal courts against California Attorney General Rob Bonta to block the enforcement of Assembly Bill 1709, a new state law designed to protect children under 16 from addictive social media features.

The law prohibits platforms from offering "psychologically exploitative features" such as personalized recommendations and autoplay to minors, and requires the deletion of children's accounts with those features. In their filings, the companies argued that the law violates the First Amendment's free speech protections and Section 230 of the Communications Decency Act. They contended that the statute wrongly compels them to adopt California's preferred methods of content presentation and substitutes the state's judgment for that of parents.

A Google spokesperson stated that overly broad laws break systems built to help kids find age-appropriate content. Meta argued that the state cannot override its editorial judgments about how to curate billions of posts. TikTok claimed that protected speech cannot be suppressed solely to shield young people from ideas deemed unsuitable by a legislative body.

The office of Attorney General Bonta said it was reviewing the lawsuits and remains committed to defending the law. Signed by Governor Gavin Newsom in September, AB 1709 is scheduled to take effect on January 1, 2027, unless a judge voids it or issues a stay. This legal challenge follows recent settlements where Meta agreed to pay up to $18 billion to resolve similar claims from nearly all US states.

California Governor Gavin Newsom has signed two key bills aimed at regulating the use of artificial intelligence in the workplace. In a review following the adjournment of the 2026 legislative session, the Electronic Frontier Foundation (EFF) highlighted the No Robo Bosses Act (SB 947) and AB 1883 as significant victories for digital rights.

SB 947 limits how employers can use automated decision-making systems to fire or discipline workers. The law ensures that an algorithm cannot be the sole authority for these consequential decisions and mandates that workers receive vital information about automated decisions affecting them. This establishes a requirement for human oversight in major employment actions in California, preventing fully autonomous AI-driven terminations or disciplinary measures.

Simultaneously, AB 1883 prohibits the use of workplace surveillance tools to predict employee emotions or collect neural data. This provision directly targets technologies that attempt to analyze psychological states, stress levels, or brainwaves through wearables or cameras. EFF argues this stops potentially biased, invasive, and exploitative forms of AI surveillance from entering the workplace, marking an important step toward protecting worker privacy and autonomy.

Beyond workplace AI regulation, the session saw other digital rights developments. SB 747 allows citizens to sue federal officials for damages if they violate the U.S. Constitution, filling gaps in remedies for First Amendment violations (such as recording police) and Fourth Amendment searches of phones and laptops. Additionally, AB 1854 and AB 1930 strengthened California's shield laws to protect medical records related to reproductive health and gender-affirming care from politically motivated out-of-state investigations.

However, not all outcomes were positive. EFF criticized the passage of AB 1709, a law banning social media use for minors under 16, arguing it cuts young people off from essential information and community without effectively solving safety issues. Furthermore, AB 1337, which sought to update the Information Practices Act to cover city and county government data handling, failed to pass the Senate, leaving a regulatory gap for local government privacy practices.

US-based artificial intelligence company OpenAI published a report on November 4, 2026, stating that it developed an autonomous vulnerability discovery, exploitation, and patching pipeline. Using an upcoming cyber model, the pipeline dynamically reproduced 1,724 vulnerabilities in the Linux kernel and generated 948 local privilege escalation (LPE) exploits and candidate patches. The total API cost for the run was $1.21343 million, averaging approximately $1,279 per LPE exploit.

OpenAI described the pipeline as consisting of five stages: vulnerability discovery, dynamic validation, exploitability assessment, patch generation, and human review. In the discovery stage, agents identified 5,034 high-value audit targets, covering about 7% of the kernel codebase. During the analysis phase, agents produced 2,308 candidate findings within 12 hours. Subsequently, validation agents reproduced 1,724 of these vulnerabilities in isolated environments, achieving a success rate of approximately 74.7%.

To assess economic impact, OpenAI compared the results with historical bug bounty programs. Cybersecurity competitions like Pwn2Own and TyphoonPWN have historically offered $20,000 to $70,000 for Linux kernel LPE exploits. Valuing the 948 exploits at these rates yields a hypothetical total between $18.9 million and $66.3 million, roughly 16 to 55 times the pipeline's cost. Additionally, in tests against Google's kernelCTF program, the pipeline discovered 21 LPE exploits for about $32,936, which would yield around $1.5 million if accepted, approximately 45 times the cost.

OpenAI also noted that the pipeline used at least 10 novel Linux exploitation techniques, which were considered new by third-party kernel exploitation experts. The company emphasized that automated patch generation is essential for securing software as model capabilities improve, urging defenders to proactively analyze their software to make exploitation cost-prohibitive for attackers.

When investors see retail sales data released each month, they rarely notice that the same month's figures are quietly revised in subsequent months. The St. Louis Fed FRED blog recently explained the mechanism behind these revisions: the U.S. Census Bureau's Annual Integrated Economic Survey (AIES) acts as a benchmark, recalibrating the levels of monthly surveys.

Monthly and quarterly surveys are designed for speed. They use smaller samples and are adept at measuring changes from one period to the next, but their estimates of absolute levels have limited precision. The AIES uses a much larger sample and more detailed information to recalibrate the levels in the monthly series after annual results are published. That is, when new annual data arrives, the faster series are revised to match the annual benchmark.

The blog provides an intuitive comparison using retail trade sales data: adding up the 12 monthly values for any year yields a total that differs from the AIES annual total by only a few million dollars. Given that annual values are in the trillions, this difference is negligible. 2024 was the latest complete annual data available at the time of writing, with both coming to approximately $7.107 trillion.

The AIES also updates the Census Business Register—the master list from which other economic surveys draw their samples. This means the annual benchmark not only revises already-published data but also affects the sample frames for future surveys.

This practice of using annual benchmarks to calibrate monthly data is not unique to the Census Bureau. The Bureau of Economic Analysis uses a similar process to derive industry output for gross domestic product (GDP) data; the Bureau of Labor Statistics uses it for the producer price index and productivity measures; and the Federal Reserve Board uses it for the index of industrial production.

In terms of coverage, FRED recently added 5,582 series from the AIES. Starting in 2023, the AIES consolidated seven independent annual business surveys into one, sent to approximately 359,000 companies. The data cover revenues, expenses, payroll, and employment for 1,046 economic activities across seven industries: agriculture, construction, manufacturing, mining, retail, services, and wholesale trade. Many series have historical data dating back well before 2023 because the Census Bureau folded results from legacy surveys into the new series: retail and wholesale back to 1992, manufacturing to 2007, and services to 2013.

For readers, the practical value of this explanation lies in understanding that once the initial monthly retail sales release is published, subsequent revisions are constrained in direction and magnitude—the annual benchmark locks the level within a relatively defined range. This helps distinguish changes in economic momentum truly reflected in monthly data from purely technical adjustments driven by sample frame and weight updates, thereby allowing for a more accurate assessment of the signal strength at the time of the initial release.

This paper, 'Towards Deep Learning Models Resistant to Adversarial Attacks,' published in 2018 by Aleksander Madry and colleagues at the Massachusetts Institute of Technology (MIT), delves into enhancing neural network security through robust optimization. The study highlights that deep neural networks have inherent weaknesses, making them highly susceptible to adversarial attacks—inputs that are nearly indistinguishable from original data but cause the model to misclassify. This vulnerability is particularly dangerous in safety-critical systems such as autonomous driving and facial recognition.

To address this issue, the authors propose a unified framework based on saddle-point (min-max) problems. This framework formalizes adversarial training as an optimization problem: the inner layer maximizes to find the adversarial perturbation that maximizes the loss function, while the outer layer minimizes by adjusting model parameters to reduce risk under this worst-case scenario. This approach not only encompasses previous attack and defense mechanisms but also provides a clear objective for safety guarantees, specifically resisting all attackers bounded by certain norms (such as the ℓ∞ norm).

In terms of implementation, the paper proves that Projected Gradient Descent (PGD) serves as a reliable 'first-order adversary' capable of effectively solving the aforementioned optimization problem. Experimental results show that models trained with PGD achieve over 89% accuracy against the strongest white-box attacks on the MNIST dataset and 46% on CIFAR-10. This indicates that by increasing model capacity and adopting correct optimization strategies, the adversarial robustness of deep learning models can be significantly improved, bringing them closer to the security standards required for practical deployment.

For researchers and engineers interested in AI safety, machine learning theory, or those needing to build high-robustness visual systems, this classic 2018 paper, 'Towards Deep Learning Models Resistant to Adversarial Attacks,' provides the foundational framework for understanding the nature of adversarial attacks and designing effective defense mechanisms. It is recommended to start reading from Section 2, 'An Optimization View on Adversarial Robustness,' to grasp the core methodology.

In the 2024 shareholder letter released by Berkshire Hathaway Inc. (United States), Warren Buffett reiterated his core communication philosophy: treating shareholders as partners. He wrote, "Our goal is to communicate with you in a manner that we would like you to use if roles were reversed—that is, if you were CEO of Berkshire and my family and I were passive investors, we would be entrusting our savings to you." This "owner mindset" is reflected not only in performance disclosures but also in the public acknowledgment of mistakes. Buffett noted that he has made errors in assessing a company's future economic prospects or management capabilities, and cited Charlie Munger to emphasize that the greatest sin is delaying the correction of errors, so-called "thumb-sucking." He criticized the boardroom culture of large listed companies that prohibit mentioning "mistakes" or "errors," arguing that this implies a false managerial perfectionism.

The letter illustrates Berkshire's unique acquisition and management philosophy through a specific case: Pete Liegl and his company, Forest River. When Buffett considered acquiring this Indiana-based recreational vehicle manufacturer in 2005, Liegl proposed an unconventional compensation plan. Liegl stated, "I looked at Berkshire's proxy statement; I don't want to earn more than my boss, so please pay me $100,000 a year." Additionally, he requested an annual bonus equal to 10% of earnings exceeding current levels. Buffett accepted this proposal, even though the parties did not explicitly define vague terms such as "appropriate" or "material." Buffett recalled, "These vague terms never caused any problems." This case demonstrates that simple incentive mechanisms are often more effective than complex contractual clauses based on high trust, and it reflects Buffett's insight into the nature of talent—he explicitly stated that he never looks at candidates' alma maters because business acumen is often innate rather than cultivated solely through elite education.

For modern readers, understanding this governance structure based on trust and long-termism is crucial. Suppose the founder of a startup faces acquisition by a large conglomerate; if the acquirer insists on cumbersome valuation adjustment mechanisms (VAMs) and strict KPI assessments, it may stifle the founder's innovative drive. Conversely, if the acquirer grants full autonomy and simple profit-sharing incentives, as Berkshire did with Liegl, it may better stimulate managerial loyalty and potential. This contrast reveals that the "human" factor in capital allocation is often more decisive than pure financial models.

The Berkshire Hathaway Shareholder Letter, written by Warren Buffett, was published in 2024. It is suitable for readers interested in corporate governance, M&A integration, and long-term investment philosophy. It is recommended to start with the chapter reflecting on "mistakes" and then delve into the case of Pete Liegl to appreciate Berkshire's unique corporate culture.