The AI Brief

The Ninth Circuit issued its opinion in No. 26-1444 on August 4, 2026, vacating Amazon's preliminary injunction against Perplexity's Comet browser and ruling Amazon is unlikely to prove Perplexity "accessed" its servers under the CFAA because users, not Perplexity, operate the tool. The court applied the rule of lenity to construe CFAA ambiguity against liability, and explicitly noted there is "little to no existing caselaw directly dealing with how to ascribe responsibility for AI agents" under the statute.

Comet, Perplexity's AI browser, requires user direction to access Amazon on its own; Amazon sent Perplexity a cease-and-desist in November 2025 after alleging the two companies had agreed to an earlier pause on agentic shopping in 2024, but Perplexity re-enabled the feature by disguising Comet's AI bot as a normal Chrome browser. Amazon filed suit in March 2026 and was awarded a temporary injunction shortly after. U.S. District Judge Maxine M. Chesney had granted that injunction on March 9, 2026, blocking Comet from accessing Amazon's logged-in account pages; the Ninth Circuit stayed the order pending appeal, heard oral arguments in Seattle on June 11, and has now vacated the injunction entirely.

The case returns to district court, where Amazon's trademark and state-law claims survive. The Ninth Circuit said its holding is narrow and that agentic AI law "will doubtless change." While the appeals court decision eliminates Amazon's injunction, it does not close the case in either company's favor; Amazon can ask for a rehearing or petition the Supreme Court.

Samsung Electronics and SK Hynix are evaluating chipmaking equipment from China's Advanced Micro-Fabrication Equipment (AMEC) for possible use at their Chinese factories, three people familiar with the matter told Reuters, as the South Korean firms hedge against the risk of tighter U.S. export controls. The memory chipmakers began testing AMEC etching equipment about two years ago, when uncertainty was mounting over whether Washington would continue allowing them to import U.S. chipmaking tools into China.

The U.S. Commerce Department designated Samsung and SK Hynix's Chinese factories as "validated end users" in 2023, allowing imports of certain controlled U.S. equipment without individual licenses. Washington revoked the VEU authorization in 2025 and later granted an annual license to the chipmakers to bring in chip manufacturing equipment to their facilities in China for 2026. Both companies remain wary that future restrictions could extend beyond new equipment to the servicing, repair, or replacement of Western tools already installed at their Chinese plants; the chipmakers are keeping Chinese suppliers in reserve as a potential way to maintain and upgrade existing production lines, rather than expand manufacturing capacity in China.

Deutsche Bank estimates that Naura Technology, AMEC, Piotech, and ACM Research will each generate more than $1 billion in revenue in 2026, and together could capture 25% to 30% of China's projected $28 billion wafer-fabrication equipment market this year; excluding lithography and metrology, Chinese suppliers' share could approach 40%. The evaluations do not establish that AMEC's tools have passed production qualification, and there is no confirmed order from either company. Chinese tools can cost 20% to 30% less than comparable equipment from established foreign suppliers, according to Dan Hutcheson, vice chair of research firm TechInsights.

OpenAI tasked GPT-5.6 with optimizing its own runtime efficiency, achieving more than 15% better token-generation efficiency through improved speculative decoding, improvements now passed on to everyone using the API, Codex, and ChatGPT. After deployment, OpenAI applied GPT-5.6 Sol to advance the frontier of efficiency by making itself more efficient to run, with results including 20% lower serving costs from production GPU kernel improvements and 15%+ better token-generation efficiency from improved speculative decoding.

For GPT-5.6 Sol, Fast mode now delivers up to 2.5× faster speeds than standard processing at twice the price, with no change in intelligence; Fast mode is backward compatible, and requests tagged "priority" will automatically use Fast mode. OpenAI also upgraded Auto-review in the ChatGPT app and Codex CLI from GPT-5.4 to GPT-5.6 Luna; combined with Luna's new pricing, Auto-review is expected to cost about 10× less.

The current model documentation carries two pricing details that matter in production: cache writes cost 1.25× the uncached input rate, and requests containing more than 272,000 input tokens are charged at 2× input and 1.5× output rates for the entire request, rules that can matter more than a headline token price in long-running agents. The Sol Fast mode launch date, per OpenAI's changelog, was July 30, 2026; Priority processing was renamed to Fast mode on that date, and speeds for gpt-5.6-sol were increased to make it 2.5× faster than standard speeds.

Mariano-Florentino (Tino) Cuéllar will join Anthropic as its first Chief Global Affairs Officer, leading the company's work on policy, strategic international engagement, and government relationships worldwide. Cuéllar will report to Anthropic President Daniela Amodei. He recently stepped down as President of the Carnegie Endowment for International Peace, a leading independent global policy research institution with scholars in 20 countries; prior to Carnegie, he was a Justice of the Supreme Court of California, where his opinions addressed technology and privacy, international agreements, and the separation of powers.

The hire comes after a turbulent year for Anthropic's government relations: the Pentagon designated the company a supply-chain risk in March after Anthropic refused to let its Claude models be used for autonomous weapons or mass surveillance of Americans, Anthropic sued the government over the blacklisting, and the case is still in court. A Trump administration export-control directive temporarily halted international sales of Anthropic's top-tier Mythos 5 and Fable 5 systems.

A special assistant in former President Barack Obama's White House, Cuéllar is tasked with finding common ground with President Donald Trump's Republican administration, which at points blacklisted and ordered controls around Anthropic's AI this year. On the same day Anthropic announced Cuéllar's appointment, White House officials met with representatives from Anthropic, OpenAI, Google, and Meta to discuss an unpublished framework to regulate their newest models and test cybersecurity protections; no results from the meeting were shared publicly. Cuéllar had served as a Trustee of Anthropic's Long-Term Benefit Trust since January 2026 and stepped down from the Trust to join the company.

Disclosure: Claude, which generates this brief, is built by Anthropic.

ChatGPT is approaching 1 billion weekly active users, according to internal information reported on July 29; OpenAI had aimed to reach the milestone by the end of 2025, putting the company about seven months behind its internal target. OpenAI reported more than 900 million weekly users and over 50 million subscribers in March 2026; the company generated $5.7 billion in first-quarter revenue but recorded a negative 122% non-GAAP operating margin.

The road to 1 billion users was a difficult one, with growth slowing last fall after user backlash to the GPT-5 model and increased competition, per The Information. Anthropic reached a $47 billion annualized revenue run rate in May, increasing pressure on OpenAI in enterprise AI. Among ChatGPT's main rivals is Google's Gemini AI, which has 950 million monthly active users and the advantage of its integration with Google Search.

The scale is difficult to separate from the cost: OpenAI generated $5.7 billion in revenue during the first quarter of 2026, but its negative 122% non-GAAP operating margin implies an operating loss of about $1.22 for every dollar of revenue, ChatGPT is becoming a global consumer platform while the economics required to serve, improve, and defend that platform remain deeply negative. OpenAI has not publicly confirmed that ChatGPT has crossed 1 billion weekly active users; the latest report says the product is nearing the threshold.