Tech news digest for Sunday, August 16, 2026

Meta's Landmark Trial Over Youth Harm Claims Begins

Meta is currently facing a significant federal trial initiated by 29 states that alleges the company's platforms, Facebook and Instagram, are designed to be addictive to young users, contributing to youth mental health issues. The outcome of this trial could lead to substantial financial penalties, potentially reaching $1.4 trillion, and enforce mandatory changes in how Meta operates its social media platforms.

• Testimony from Meta CEO Mark Zuckerberg and Instagram CEO Adam Mosseri is anticipated during the trial.
• A New Mexico court recently ordered Meta to pay $375 million in civil penalties related to similar claims about youth mental health.

'Spending is leading to earnings': Wall Street strategists see payoff from Big Tech's AI investment

Wall Street strategists are observing that substantial investments in AI by major tech companies are beginning to yield earnings, particularly for hyperscalers like Microsoft, Amazon, Alphabet, and Meta. Keith Lerner, Truist's chief investment officer, noted that the spending is translating into earnings, with Microsoft reporting record cloud revenue as its Azure business surpassed $100 billion in annual sales. Amazon's AWS experienced a growth rate of 36.7%, the fastest in 18 quarters, while Alphabet's cloud business also reported significant growth. Despite anticipated negative free cash flow for FY27, demand is catching up with capital expenditures, indicating a potential for stronger revenue growth ahead.

• JPMorgan raised its S&P 500 price target to 8,000 from 7,800.

Top mathematicians say LLMs are strong calculators but poor creative thinkers

Mathematicians Timothy Gowers and Peter Sarnak have praised large language models (LLMs) for their mathematical capabilities but caution against their limitations in generating original ideas. Gowers points out that while LLMs can explore multiple search paths and combine existing methods, they lack the intuition necessary to identify the most promising routes. Sarnak adds that these models can derive results from established theories but struggle to create the underlying abstractions needed for significant proofs from basic questions. This assessment aligns with findings from DeepMind researcher Tom Zahavy, who attributes these shortcomings to the models' inability to perform 'manipulative abduction,' a process essential for inventing new foundational assumptions.

Bonta Says Meta Put Profit Over Children’s Safety

California Attorney General Rob Bonta has launched a landmark youth safety case against Meta, alleging the company prioritized profits over children's safety by knowingly designing harmful features on Facebook and Instagram. Bonta characterized the case as a 'tobacco moment,' highlighting the potential long-term impacts of social media on youth. The state is seeking financial penalties and demands changes to Meta's products, while Meta refutes the claims, stating it has taken steps to enhance the safety of its platforms for young users.

• Meta denies the allegations, claiming it has implemented measures to enhance safety for young users.

OpenAI dissolved the team built to catch catastrophic AI risks, reassigning its work to other groups

OpenAI has dissolved its 'Preparedness' team, which was responsible for assessing serious and catastrophic risks associated with its AI models, as of July 2026. The team's evaluations of biological and cyber risks have been reassigned to other existing teams. Dylan Scandinaro, the former unit lead, is now concentrating on safety risks from 'recursively self-improving' AI. This organizational change has led to recent departures, including Chief Ethics Officer Chloe Bakalar, and has sparked internal unease regarding the company's commitment to safety, especially following a recent autonomous hacking incident involving Hugging Face.

• Internal sources report growing unease about safety measures at OpenAI.

Anthropic's bio-weapons filter was down for nearly a year, exposing 133 million requests

Anthropic's bio-weapons filter, designed to prevent AI misuse related to chemical and biological weapons, was inactive from May 2025 through April 2026, exposing 133 million requests. During this period, around 50,000 external contractors interacted with the AI models without the necessary safety measures in place. Although Anthropic conducted an internal investigation and found no evidence of misuse, the company has since tightened its contractor vetting processes to enhance safety and compliance.