Five AI stories from the week of April 13–April 19, 2026 that actually mattered — what changed, why it matters, and where to read more.
In this issue
- Stanford HAI publishes ninth annual AI Index Report
- OpenAI releases GPT-5.4-Cyber under tiered KYC access program
- Anthropic ships Claude Opus 4.7 with large coding and vision gains
- OpenAI expands Cerebras compute deal to more than $20 billion
- Cerebras files S-1 for NASDAQ IPO, discloses $510M 2025 revenue
Stanford HAI
Stanford HAI publishes ninth annual AI Index Report
Stanford’s Institute for Human-Centered Artificial Intelligence released its 2026 AI Index on April 13, documenting that generative AI has reached 53% population adoption—faster than the PC or the internet—while organizational adoption hit 88%. The report also found that responsible-AI practices are falling further behind capability growth, with documented incidents rising from 233 to 362 year-over-year, and that the performance gap between U.S. and Chinese frontier models has effectively closed.
OpenAI
OpenAI releases GPT-5.4-Cyber under tiered KYC access program
OpenAI published GPT-5.4-Cyber on April 14, a fine-tuned variant of GPT-5.4 with lower refusal boundaries for defensive cybersecurity tasks including binary reverse engineering without source code. Access is restricted to identity-verified security vendors, organizations, and researchers through OpenAI’s Trusted Access for Cyber program, with no public availability; the model is described as ‘cyber-permissive’ and represents the first time OpenAI has shipped a deliberately capability-expanded model to a restricted audience.
Anthropic
Anthropic ships Claude Opus 4.7 with large coding and vision gains
Anthropic made Claude Opus 4.7 generally available on April 16, reporting a 13% lift on a 93-task coding benchmark, 3x more production tasks resolved than Opus 4.6, and a jump in visual-acuity benchmarks from 54.5% to 98.5% through support for images up to ~3.75 megapixels. The release also adds a new ‘xhigh’ extended-thinking effort level and ships at the same API pricing as Opus 4.6 ($5 / $25 per million input/output tokens), with availability across the API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry.
The Information (via GuruFocus)
OpenAI expands Cerebras compute deal to more than $20 billion
The Information reported on April 17 that OpenAI has agreed to pay Cerebras more than $20 billion over three years for wafer-scale chip infrastructure, doubling a previously reported $10 billion commitment from January and adding approximately $1 billion toward Cerebras data center buildout. OpenAI will receive equity warrants that could reach a ~10% stake as spending milestones are hit, representing a significant diversification away from NVIDIA as the sole inference compute supplier.
Axios
Cerebras files S-1 for NASDAQ IPO, discloses $510M 2025 revenue
Cerebras Systems filed an S-1 registration statement on April 18, targeting a NASDAQ listing under ticker CBRS after withdrawing a prior filing in October 2025 due to a national security review. The S-1 discloses 2025 revenue of $510 million (76% year-over-year growth) and a GAAP net income of $237.8 million, though the company reported a non-GAAP net loss of $75.7 million; the OpenAI compute deal is identified as a central commercial relationship underpinning the offering.