Memory Now Free for All Users — Plus a ChatGPT Import Tool
Anthropic made its memory feature available to free users starting Monday, removing what had been a paid-only perk since October 2025. Memory lets Claude recall preferences and past interactions across conversations to deliver more personalized responses over time. The announcement came as Claude continues to post all-time records for daily sign-ups following the Pentagon dispute.
Alongside the free memory rollout, Anthropic launched a cross-AI import tool that lets users pull conversation history and saved memories from ChatGPT, Google Gemini, and Microsoft Copilot into Claude with a simple copy-paste. The goal: eliminate friction for users switching platforms. New arrivals don't start from scratch anymore.
Structured Outputs Now Generally Available on the Claude API
Structured outputs — JSON schema enforcement for Claude's responses — have graduated to GA across Claude Sonnet 4.5, Opus 4.5, and Haiku 4.5. Developers no longer need to include a beta header in their requests. This release also brings expanded schema support, significantly improved grammar compilation latency, and a simplified integration path via the output_config.format parameter. Reliable, schema-validated output is now a first-class API primitive.
Data Residency Controls Launch — US-Only Inference Now Available
Anthropic launched new data residency controls via the inference_geo parameter in the Messages API, letting enterprise customers specify where model inference runs. US-only inference is available at a 1.1x price premium for all models released after February 1, 2026. The feature addresses a key blocker for regulated industries — healthcare, finance, and government — where data locality is a legal or compliance requirement, not a preference.
Claude Code Analytics API Now Live for Organizations
Anthropic launched the Claude Code Analytics API, giving organizations programmatic access to daily aggregated usage metrics for their Claude Code deployments. Data is aggregated per organization, per day, and each endpoint returns a snapshot for a single specified date. For teams managing large Claude Code rollouts, this is the missing piece for cost tracking, productivity benchmarking, and capacity planning — no more manual exports from the console dashboard.
Fine-Grained Tool Streaming Goes GA Across All Models
Fine-grained tool streaming is now generally available on all Claude models and platforms, with no beta header required. Developers building agentic pipelines can now stream tool call inputs in real time as they're generated, rather than waiting for the complete payload. This meaningfully improves perceived latency for tool-heavy workflows — especially long-running agent chains where tool calls are frequent and input payloads can be large.
Cowork Gets Scheduled Tasks and a New Customize Hub
The latest Claude Desktop and Cowork release introduces scheduled tasks — users can now create recurring or on-demand automations directly in Cowork, picking cadence from daily, weekly, weekday, hourly, or on-demand. A new Customize section in Claude Desktop consolidates skills, plugins, and connectors in one place. Setup is either through the /schedule command mid-task or from the new Scheduled sidebar panel. This is Cowork's biggest workflow upgrade since launch.
Amodei to CBS: "We Have Red Lines" — Pentagon Showdown Heads to Court
In an exclusive CBS News interview, Anthropic CEO Dario Amodei made clear the company is not backing down from its ethical constraints on military AI use. "We are still interested in working with them as long as it is in line with our red lines," he said, citing Anthropic's refusal to allow Claude to be used for fully autonomous weapons or mass domestic surveillance of Americans. Anthropic has confirmed it will challenge the Pentagon's "supply-chain risk" designation in court.
Legal analysts at Lawfare argue the designation is unlikely to survive judicial review, noting procedural irregularities and First Amendment concerns around compelled speech in commercial contracts. Meanwhile, OpenAI announced a Pentagon deal that legal scholars say reflects exactly the kind of compromise Anthropic refused — raising questions about what principles, if any, constrain AI companies in national security contexts.
The Outage Aftermath: Enterprise AI Reliability Under the Microscope
Yesterday's 2-hour, 45-minute worldwide Claude outage — affecting claude.ai, the console, and Claude Code while the API stayed live — is driving a broader conversation about enterprise AI reliability. Windows Forum and VPN Central both published post-mortems noting that as enterprises lean harder on Claude for mission-critical tasks, single-vendor auth infrastructure failures are becoming a serious risk category. The timing was particularly damaging: the outage hit as Anthropic was riding a record wave of new user sign-ups, giving competitors a moment to pounce. Anthropic has confirmed services are fully restored and monitoring is ongoing.
OpenAI's Pentagon Deal Is Exactly What Anthropic Refused
Hours after Anthropic was declared a supply-chain risk, OpenAI announced a deal to deploy its models within the Defense Department's classified network. MIT Technology Review's analysis argues the OpenAI agreement grants the exact capabilities Anthropic refused to enable — including use cases that approach the autonomous-weapons and surveillance boundaries Anthropic drew in its original contract. The contrast is fueling intense debate in AI ethics circles and on Hacker News about the long-term consequences of competitive pressure on AI safety commitments.
Haiku 3 Retirement Countdown: April 19 Deadline Looms
A reminder that's picking up urgency in developer communities: Claude Haiku 3 (claude-3-haiku-20240307) retires on April 19, 2026. Any application still hitting that model string will break after the deadline. The migration path is Claude Haiku 4.5, which delivers meaningfully improved performance at comparable pricing. With 47 days left, teams with large Haiku 3 deployments should be testing migration builds now — not the week before retirement.