The AI Brief

Google confirmed its Made by Google 2026 hardware event for August 12 in New York, with the keynote beginning at 6 p.m. ET. The event is expected to unveil five devices, Pixel 11, Pixel 11 Pro, Pixel 11 Pro XL, Pixel 11 Pro Fold, and Pixel Watch 5. Gemini 3.5 Pro has been the central AI question mark attached to the event.

The model was announced at Google I/O on May 19, 2026, with a "next month" rollout target, which slipped past June and July. Bloomberg reported on July 16 that the release was delayed again after the model fell short of Google's own internal quality goals, reportedly on hallucination rates and real-world reliability, with DeepMind said to have scrapped and rebuilt the base model. As of August 8, Gemini 3.5 Pro remained unavailable to the general public, with DeepMind's model pages still listing it as "coming soon" and API changelogs showing only the July 21 releases of 3.6 Flash and 3.5 Flash-Lite.

As of August 8, a new rumor pointed to August 12, again from finance and insider sources, still unconfirmed by Google, carry it with the same caution as the missed July rumor. Leaks cite a 2-million-token context window, an upgraded Deep Think reasoning layer, and significantly improved agentic coding performance. Enterprise partners have already been quietly testing early builds, and eagle-eyed users spotted unnamed contenders on LM Arena. No model card, pricing, or API model ID had been officially published as of this edition's research window. The brief will update if the launch is confirmed before publication; if not, the next edition carries the outcome.

First covered in Vol. I, No. 76 (August 9, 2026) as the live window for Gemini 3.5 Pro.

OpenAI announced Premium seats for ChatGPT Business on August 10, offering five times more usage than Standard seats and removing the five-hour usage limit, so teams can take on larger projects with fewer interruptions within the same secure Business workspace. Premium seats cost $125 per user per month when billed monthly, or $100 per user per month billed annually; Standard seats remain at $25 per user per month.

The new seat fills a gap that has existed since OpenAI rebranded ChatGPT Team as ChatGPT Business in August 2025. Before the announcement, business customers who had outgrown Standard seats faced a binary choice: stay throttled, or negotiate a full Enterprise contract with custom pricing, dedicated support, and a full sales cycle. Enterprise contracts typically require direct engagement with OpenAI's sales team and carry six-figure annual commitments.

OpenAI said the change reflects how teams are tackling more complex tasks with ChatGPT and need more capacity, put plainly, agent-based AI burns through far more tokens. Both seat types can be mixed within the same workspace, letting admins assign the right tier to each team member. Starting August 19, 2026, OpenAI is also changing when charges for additional paid seats occur, instead of appearing on a later bill, charges post immediately when a seat is added. OpenAI is offering up to $100 in workspace credits per Premium seat (up to five seats) for workspaces that join the waitlist before August 20.

OpenAI made Daybreak capabilities available through Amazon Bedrock on August 11. With Daybreak Access, defenders can use frontier cyber models within their existing AWS environments. Daybreak Blue provides access to frontier general-purpose models, including GPT-5.6 Sol, with safeguards tailored to authorized defensive security work. Daybreak Red provides access to OpenAI's purpose-trained cybersecurity models for authorized vulnerability research, exploit validation, and security testing.

According to OpenAI, security researchers used GPT-5.6 Cyber through Daybreak Red to identify two previously unknown vulnerabilities in V8, the JavaScript engine used by Chrome; when chained together, they could enable memory corruption and a heap sandbox escape. The initial vulnerability was fixed and released as CVE-2026-15903. Both models run on Amazon Bedrock's next-generation inference engine; zero-operator access is enforced at the chip, so even AWS operators cannot access prompts and completions during inference.

Daybreak Red and Daybreak Blue run in US East (N. Virginia), and access requires enrollment in OpenAI's Trusted Access for Cyber vetting program. OpenAI recommends Blue as the starting point for most security teams; Red carries a lower refusal threshold matched by stronger identity verification, monitoring, and access controls, and is aimed at authorized vulnerability research and exploit reproduction. The Bedrock placement means qualified customers can access the models through familiar AWS security and governance workflows, and usage counts toward existing AWS cloud commitments.

On August 10, Anthropic, Macquarie Asset Management, and GIC announced a strategic partnership to establish Theseus Infrastructure, a new platform to develop, operate, and lease data center infrastructure at scale to Anthropic under long-term agreements. The parties will work together to identify and develop new sites for which Anthropic will be the anchor tenant, with an initial focus on the United States; funds managed by Macquarie Asset Management will, together with GIC, own the platform and fund the majority of the equity for each project.

Alongside the venture, Anthropic pledged to pay 100% of grid-upgrade costs and cover consumer electricity price increases tied to its data center demand. The companies did not share any details on planned spending or the size of the projects. GIC, Singapore's sovereign wealth fund, is an existing Anthropic investor with prior data center experience across Equinix xScale, EdgeCore, and Vantage platforms; Macquarie has backed Netrality, Applied Digital, and Virtus in similar infrastructure roles.

Anthropic has made relatively few splashy partnerships for data center projects compared to rival OpenAI, which launched its Stargate infrastructure joint venture. Theseus follows Anthropic's $9.1B deal with Bitcoin miner Riot Platforms (covered August 11) and a $10B deal with Volta Infra (covered August 7), but where those are capacity leases, Theseus creates a permanent institutional vehicle. The strategic partnership follows a commitment Anthropic made last year to spend $50 billion to build its own AI data center campuses across the United States.

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

On August 11, ChatGPT Ads launched in the United Kingdom, Mexico, Brazil, Japan, and South Korea, with OpenAI continuing to expand to more markets this year. The move extends a program that began in the United States in early 2026 and subsequently covered Canada, Australia, and New Zealand. Ads do not influence the answers ChatGPT gives; answers are optimized based on what is most helpful to the user. When ads appear, they are clearly labeled as sponsored and visually separated from the organic answer.

During the test, OpenAI decides which ad to show by matching ads submitted by advertisers with the topic of the user's conversation, past chats, and past interactions with ads, for example, a user researching recipes may see ads for meal kits or grocery delivery. Users paying for Plus, Pro, Business, and Enterprise tiers do not see advertisements.

OpenAI has not published an updated annualized revenue figure since a reported roughly $25 billion run-rate figure earlier in 2026. The ad expansion lands as OpenAI's public S-1 is expected on SEC EDGAR in late August, the first filing to disclose audited financials including what the advertising line contributes. OpenAI opened self-serve ad buying through its Ads Manager platform in May 2026, with cost-per-click bidding; early campaigns ran for Best Buy, Lowe's, and VistaPrint. The five-market expansion adds two G7 economies (UK, Japan) and the largest Latin American market (Brazil) to the addressable ad inventory before the company's roadshow.