Volume 343, Issue 2 · 11 articles from 3 issues
Anthropic's prospectus shows heavy revenue concentration, giving enterprise CIOs more negotiating leverage with AI vendors.
CIO.com reports that, per a Reuters analysis of Anthropic's confidential prospectus, nearly a quarter of the company's revenue last year came from two customers and many large clients lack long-term contracts.
Analyst Frank Dickson says this gives CIOs more leverage than they realize, urging them to negotiate price protection, notice before model retirements and data portability, and to avoid hardwiring one model into every workflow by using an abstraction layer, owned evaluations and portable prompts and data.
io.net CMO Jack Collier says the filing reads like a case study in concentration risk.
Anthropic releases Claude Sonnet 5.5, a faster, cheaper upgrade over Sonnet 5 that complements Opus 5.5.
Anthropic introduced Claude Sonnet 5.5, the second model in the Claude 5.5 family, which runs more than 30% faster than Sonnet 5 and costs up to 30% less for most work.
Sonnet 5.5 targets well-scoped everyday tasks, bug fixing and polished documents, slides and spreadsheets, and scores 70.6% on the Terminal-Bench 4.0 agentic coding evaluation.
Claude Haiku 5.5 is expected to join the family in the coming weeks.
Anthropic launched Claude Haiku 5.5, its cheapest and fastest small model, at about 75% lower cost than Haiku 4.5.
Anthropic released Claude Haiku 5.5, billed as its cheapest, fastest and most capable small model for high-volume, cost-sensitive work such as summaries, compaction, database queries and classification.
The model works well as a subagent alongside Opus 5.5 and Sonnet 5.5 on coding tasks, suits live customer support and browser use, and costs about 75% less to run on average than Haiku 4.5.
It is also the first Haiku-class model with an adjustable effort setting, and Anthropic separately halved Sonnet 5.5 cache-read prices and added a monthly API credit for Claude Max and Team subscribers.
Anthropic launched its Cyber Mission, starting with a critical infrastructure defense program and a free OSS Scanner for open-source projects.
Anthropic announced the Cyber Mission, a long-term effort to give defenders tools, research and resources to secure the systems everyone depends on, citing state-sponsored footholds across sectors and resource shortages among defenders.
Its new Critical Infrastructure Defense Program brings frontier models, on-site engineers and threat research to defenders of power grids, water systems, transportation networks and government systems.
For open-source software, Anthropic launched OSS Scanner, which gives open-source projects free regular security scans from its strongest models.
Anthropic expanded its Cyber Verification Program into three access tiers that give vetted defenders reduced cyber safeguards.
Anthropic has expanded its Cyber Verification Program, giving qualifying security professionals access to advanced cyber capabilities with reduced blocking classifiers.
The program now offers three access tiers covering its most capable models, including Claude Opus 5.5, Claude Sonnet 5.5 and Claude Mythos 5.1, while generally available models keep conservative safeguards because cybersecurity is dual use.
It merges Project Glasswing, which gave critical-software organizations access to Claude Mythos, with the original CVP, which reduced safeguards for vetted teams.
Claude Code Review exits research preview for Team and Enterprise, with reviews about 3x faster and better bug detection.
Anthropic's Code Review is out of research preview and available to Claude Team and Enterprise organizations, with eligible Enterprise customers able to claim at least $10,000 in credits by January 15, 2027.
For general availability, Anthropic made reviews about 3x faster, cutting typical times from about 20 minutes to 6-10 minutes.
It also improved detection of bugs that matter and made reviews quieter, sized to each change and easier to tune per repository.
Give Claude Opus 5.5 a task, labeled screenshots and settled constraints, then verify what screenshots can't show.
The guide explains how to get actionable UI design feedback from Claude Opus 5.5's vision capabilities.
Start by stating what the user is trying to finish, supply labeled screenshots of each state with viewport sizes and a full-screen view alongside close-ups, and follow Anthropic's advice to use clear, uncompressed images.
Say which decisions are settled, and ask for a short list of findings that each name the element, the obstacle and a way to confirm it, keeping observed evidence separate from preferences.
Because screenshots can't show keyboard navigation, validation or responsiveness, and the model's coordinate estimates are approximate, verify findings on the working interface.
Uber CISO Philip Martin explains why AI agents need an insider threat program and how Claude supports Uber's threat modeling.
Uber CISO Philip Martin leads security for a platform handling over 40 million trips and delivery orders a day.
His team began using Claude for threat modeling before Claude Code existed, and Claude now gathers context itself from code, design docs and ticket history.
Martin argues AI agents need the same kind of insider threat program companies run for employees, with a focus on accidental misuse, and describes how Uber's open-source monitoring tools work alongside Anthropic's Enterprise Frontier Safeguards.
The Claude Code team shifted most code reading to machines, doubling merged PRs while cutting reverts by over 40%.
Claude writes about 90 percent of Claude Code's own code, so the team redesigned code review to minimize human review time without raising reverts or incidents.
From June to September, merged pull requests more than doubled while the share reverted fell by more than 40 percent.
The team moved most reading to machines, optimized for human attention and invested in making mistakes fast and easy to undo, and the post walks through each review layer and how other teams can adopt the process.
Claude Managed Agents, the Messages API and Claude SDKs can now send OpenTelemetry traces to any OTel observability tool.
Developers can now send OpenTelemetry traces from Claude Managed Agents, the Messages API and Claude SDKs to any OTel-compatible tool, tracing each model call, tool use and MCP server call.
Destinations are added in the Claude Console, and traces include latency, tokens and IDs.
Because they follow OpenTelemetry GenAI semantic conventions, Claude traces work alongside data from other providers in tools such as Grafana Cloud, Honeycomb, Sentry, Splunk, Datadog, Langfuse and LangSmith.
Anthropic built Scion, an in-house version control system, after daily merged changes grew from 2,500 to over 30,000.
With Claude now iterating autonomously on code, Anthropic says developer infrastructure built for human scale, such as source control, builds and CI, has become the bottleneck.
Merged code changes rose from about 2,500 per day in May 2026 to over 30,000 per day by September.
To keep up, Anthropic built Scion, an in-house version control system, and migrated its monorepo off git in five months, alongside a code review surface called Antique, code search and a source control IAM system.