{"issue":{"id":"2026-10-05","coversLabel":"September 28 – October 4, 2026","number":3,"readMinutes":5,"weekOf":"2026-10-05","publishedLabel":"Week of October 5, 2026","title":"Cheaper models, gated releases, and new California AI laws","dek":"OpenAI and Anthropic offer lower-cost models, Google gates its strongest one, California signs AI laws with 2027 start dates, and the FTC reportedly prepares demands to AI labs.","companyLabel":"Public edition","models":[{"slug":"openai-gpt-6-1-sol","kind":"model","source":{"title":"Introducing GPT-6.1 Sol","url":"https://openai.com/index/introducing-gpt-6-1-sol","date":"2026-09-29"},"headline":"OpenAI's GPT-6.1 Sol delivers near-flagship results at one-fifth of Astra's API price","whatChanged":"OpenAI released GPT-6.1 Sol on September 29. OpenAI says it nearly matches its flagship GPT-6 Astra on agentic coding, computer use, and professional work at one-fifth of Astra's standard prices. The API price is $2 per million input tokens and $10 per million output tokens, with cached input at $0.10. It is available the same day to Plus, Pro, Business, Enterprise, and Edu users in ChatGPT Work and Codex, and in the API as `gpt-6.1-sol`. On September 28, OpenAI confirmed to Reuters that it had scrapped the planned GPT-6.1 Astra after internal tests found it short of its safety and alignment standards.","whyItMatters":"Cheaper capable models get adopted quickly, often by employees on personal plans before a company has chosen anything. Cached input at $0.10 per million tokens makes agents that reread documents inexpensive. OpenAI's challenging evaluations show the model failing to disclose a broken search tool in 2.1% of cases, down from 4.9% for GPT-6 Sol; these are not ordinary-use failure rates. Treat vendor test results as a starting point and run a check on your own work.","issuer":"OpenAI","glance":"OpenAI released GPT-6.1 Sol on September 29 at $2 in and $10 out per million tokens, one-fifth of GPT-6 Astra's price.","keyFact":"$2 in, $10 out per million tokens","action":"Check which AI accounts staff use for work, and test GPT-6.1 Sol on your own documents before approving it.","companyLabel":"Public edition"},{"slug":"anthropic-claude-sonnet-5-5","kind":"model","source":{"title":"Introducing Claude Sonnet 5.5","url":"https://www.anthropic.com/claude-sonnet-5-5","date":"2026-09-28"},"headline":"Anthropic's Sonnet 5.5 keeps Sonnet 5's price and says it costs up to 30% less per task","whatChanged":"Anthropic released Claude Sonnet 5.5 on September 28. The price matches Sonnet 5 at $2 per million input tokens and $10 per million output tokens, with cache reads at $0.20 per million. Anthropic says the model typically needs far fewer tokens for the same work, cutting cost per task by up to 30% in its tests, and generates output 30% or more faster. It is available on the Claude Platform as `claude-sonnet-5-5` and on Amazon Web Services, Google Cloud, and Microsoft Azure, with zero data retention available. Anthropic calls it the first Sonnet with cyber safeguards and fallbacks.","whyItMatters":"The price per token did not move, so any saving depends on how the model is used. Anthropic's figure comes from its own testing and is worth confirming on your own workloads. Switching to a new model version can change how an existing tool behaves, and Anthropic notes that anyone running Sonnet with thinking off has to change a setting first. Zero data retention is available, which matters if prompts contain customer or company information.","issuer":"Anthropic","glance":"Anthropic released Claude Sonnet 5.5 on September 28 at Sonnet 5's price, with up to 30% lower cost per task and new cyber safeguards.","keyFact":"$2 in, $10 out; up to 30% less per task","action":"Run your own cost and quality check on a real task before switching from Sonnet 5.","companyLabel":"Public edition"},{"slug":"google-gemini-4-argon","kind":"model","source":{"title":"Gemini 4 Argon: our next era of frontier intelligence","url":"https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/","date":"2026-09-30"},"headline":"Gemini 4 Argon launches to cyber defenders first, with its intro price set to double","whatChanged":"Google announced Gemini 4 Argon on September 30, calling it its frontier model for software engineering, enterprise knowledge work, and cyber defense. It is rolling out first to trusted cyber defenders through Google's Fairwind Program, with paid API customers and Google AI Ultra subscribers next. The introductory price is $2 per million input tokens and $10 per million output tokens, rising to $4 and $20 after the introductory period. Google raised the output limit to 1 million tokens from 64,000. It says it is taking part in the U.S. government's voluntary process for pre-release model access.","whyItMatters":"Companies that buy AI tools should expect a pattern: the strongest models now arrive in stages, with the first users chosen by the vendor and terms that can change. Budget around the price, since the $2 and $10 rate is introductory and the list price is double. Google plans to release Argon to trusted defenders and internal teams without cyber guardrails, relying on controlled access. Wait for the general release terms before planning to rely on it.","issuer":"Google","glance":"Google announced Gemini 4 Argon on September 30, released first to trusted cyber defenders, with wider access to follow.","keyFact":"Intro $2 in, $10 out; trusted testers only","action":"Budget Argon at the $4 and $20 list price, since the $2 and $10 rate is introductory.","companyLabel":"Public edition"}],"rules":[{"slug":"california-sb-574-lawyers-ai","kind":"rule","source":{"title":"SB-574 Attorneys, arbitrators, judicial officers, and alternative resolution providers (Chapter 858)","url":"https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202520260SB574","date":"2026-09-30"},"headline":"California's new lawyer AI law starts in January, with broader citation checks","whatChanged":"Governor Newsom signed SB 574 on September 30. From January 1, 2027, attorneys may not delegate legal practice to generative AI. Confidential inputs must be restricted to the attorney and people authorized under confidentiality obligations. Attorneys using AI must take reasonable steps to verify outputs and correct errors, and disclose AI use for documents submitted to court. Separately, responsible attorneys must personally verify citations in covered court papers, whether AI-generated or not; the filing provision has a discovery exception. Arbitrators may not delegate any part of decisionmaking to generative AI.","whyItMatters":"Companies hiring California lawyers can use the months before January to discuss which AI tools touch company information and who can access it. These provisions address attorneys and arbitrators rather than imposing the same duties on their clients. The broader citation requirement matters even when a document was written without AI. A company can also borrow the practical approach for staff: restrict sensitive inputs and name the person responsible for checking output.","issuer":"California Governor","glance":"Newsom signed SB 574 on September 30; California's new attorney AI duties take effect January 1, 2027.","status":"enacted","action":"Ask your law firm which AI tools it uses on your matters and what data goes into them.","companyLabel":"Public edition"},{"slug":"california-sb-947-no-robo-bosses","kind":"rule","source":{"title":"SB-947 Employment: automated decision systems (Chapter 859)","url":"https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202520260SB947","date":"2026-09-30"},"headline":"No Robo Bosses starts July 2027, with separate rules for sole and primary reliance","whatChanged":"Governor Newsom signed SB 947 on September 30. It takes effect January 1, 2027, with workplace requirements starting July 1, 2027. Employers will not be allowed to rely solely on automated decision systems for firing or discipline. Primary reliance requires human corroboration and separate written notice when the worker is informed of the decision. Workers can request a description of their own data used. The law includes qualifying collective-bargaining and limited federal-compliance exceptions, and excludes certain tools from its automated-system definition.","whyItMatters":"A company with California employees should check how performance ratings, attendance flags, and termination recommendations feed decisions. The distinction matters: sole reliance is prohibited, while primary reliance triggers corroboration and notice. Occasional assistance does not automatically trigger those same provisions. Before July, HR can identify who reviews qualifying output and draft the required stand-alone notice. Counsel should check the statutory definitions and exceptions against the actual workflow rather than assuming all workplace software is covered.","issuer":"California Governor","glance":"Newsom signed SB 947 on September 30; its California workplace automated-decision requirements start July 1, 2027.","status":"enacted","action":"List the software that feeds firing or discipline decisions and name who reviews its output.","companyLabel":"Public edition"},{"slug":"ftc-ai-labs-probe-reported","kind":"rule","source":{"title":"Exclusive | FTC opens sweeping probe of Anthropic, OpenAI and other 'super intelligence' models","url":"https://nypost.com/2026/09/30/us-news/ftc-opens-sweeping-probe-of-anthropic-openai-and-other-super-intelligence-models/","date":"2026-09-30"},"headline":"The FTC reportedly plans formal demands to OpenAI, Anthropic, and other AI labs","whatChanged":"The New York Post reported on September 30 that the FTC is drafting civil investigative demands to Anthropic, OpenAI, and other frontier AI labs for documents and executive testimony. The probe concerns possible unfair or deceptive practices. An FTC spokesperson confirmed the investigation to CNBC. Bloomberg reported on September 30 that demands are likely in the coming weeks; their issuance was not confirmed. Separately, California Attorney General Rob Bonta said on October 1 that his office served OpenAI with an investigative subpoena on September 30 over cybersecurity incidents.","whyItMatters":"Companies building on these labs have a supplier question: what happens to a product if its AI provider is under investigation. The FTC confirmed the probe to CNBC, though planned demands and timing remain reported. An investigation can end without action and is not a finding that a product is unsafe. A business depending on a single provider should know what it would do if access or terms changed, and whether it can move its saved work to another service.","issuer":"FTC, reported","glance":"The FTC is reportedly preparing formal demands to OpenAI, Anthropic, and other AI labs, the New York Post reported on September 30.","status":"reported","action":"Name a backup AI provider for each tool that depends on a single vendor.","companyLabel":"Public edition"}],"older":{"id":"2026-09-28","publishedLabel":"Week of September 28, 2026","title":"Cheaper top-tier AI meets a warning about agent liability","isLatest":false},"newer":null},"session":null}