Pentagon Hands Classified-Network AI Contracts to Seven Vendors — Anthropic Stays Blacklisted, Mythos Carved Out as a “National Security Moment”
The DOD awarded classified-network AI contracts on May 1 to OpenAI, Google, Microsoft, Amazon Web Services, Nvidia, SpaceX, and the startup Reflection AI. Anthropic was the one frontier lab left off the list. The exclusion stems from the supply-chain-risk designation Defense Secretary Pete Hegseth formalized in March, which the company is fighting in court — Anthropic sued the Trump administration in March to try to reverse the blacklisting. The core dispute traces back to contract renegotiations earlier this year: the department demanded Claude be available for “all lawful purposes,” while Anthropic insisted on walling off mass domestic surveillance and the development of autonomous weapons. The procurement question is now resolved for everyone except Anthropic; the safety-policy question that produced the impasse is the same one the EU and UK negotiations are now circling.
The Mythos track is the wrinkle. DOD CTO Emil Michael said Friday that Anthropic remains a supply-chain risk, but Mythos is a “separate national security moment.” The NSA is already running Mythos Preview despite the Pentagon’s own blacklisting, and senior White House officials including Dario Amodei met earlier this month to discuss expanded Mythos access. President Trump told CNBC after that meeting that “it’s possible” a deal between Anthropic and DOD lands. The structural read going into the week: the IPO process now has a clean private-procurement narrative for everyone outside the federal supply chain (the $1.5B Wall Street JV announces inside the same window), a separate Mythos-only federal track, and a still-unresolved blacklist on the conventional contract. Watch the Eurogroup readout today and any movement on the Trump-administration federal-onramp executive action this week for the next data points.
$1.5B Wall Street Joint Venture Nears Close — Blackstone, Goldman Sachs, Hellman & Friedman Anchor an AI Consulting Arm for PE Portfolios
WSJ reporting confirmed by Reuters and the AP wire: Anthropic is finalizing an approximately $1.5B joint venture with Blackstone, Goldman Sachs, and other Wall Street firms to sell AI tools into private-equity-backed companies. Anthropic, Blackstone, and Hellman & Friedman are anchoring the deal at roughly $300M each; Goldman Sachs is a founding investor at around $150M; General Atlantic and additional firms round out the total to about $1.5B. The vehicle will operate as a consulting arm for Anthropic, helping PE firms’ portfolio companies integrate Claude into finance, operations, customer service, analytics, and enterprise software. Announcement timing per the reporting: this week. The structural similarity to the Accenture and Infosys multi-year partnerships is intentional — Anthropic is now stitching together a vertical-by-vertical channel network on top of the Claude Platform.
Read for the IPO clock: this is the first piece of news that lands cleanly inside the May calendar after the Friday-night $40–50B at $850–900B raise news, and the timing reads as deliberate. The funding round mark sets the private-market valuation; the Wall Street JV gives the S-1 process a financial-services consulting line on top of the API revenue ramp. The Briefing: Financial Services event Anthropic is hosting tomorrow May 5 was already on the calendar — the JV announcement could plausibly fold into that surface, or land independently in the next 48 hours. Either way, the timing puts a financial-services revenue narrative on the table the same week as the Eurogroup-Mythos discussion, the Code with Claude SF developer conference, and any Sonnet 4.8 ship. The S-1 is now picking up a vertical channel narrative every 7–10 days.
Eurogroup Takes Up Mythos with Bank Supervisors Today — Germany Drives the Access Push, France and Italy Expected to Align
The discussion previewed Sunday is happening today: euro-area finance ministers meet bank supervisors at the Eurogroup with Mythos access on the agenda. Germany’s BaFin chief Michael Theurer is the named driver, demanding the European Commission obtain access either from Anthropic or via the US administration; France and Italy are reported aligned with the German posture. The European Banking Authority and the EU AI Office have not pre-staked public positions — today is their first formal inflection. The reference architecture being held up is the British framework: AI Security Institute on the testing side, FCA and NCSC on the regulatory side, named bank participants under Glasswing-equivalent terms. Anthropic has not commented publicly. Watch the readout language for whether “UK-equivalent terms” or a distinct “EU framework” surfaces as the working phrase.
The political question for Brussels: ask for parity with the UK arrangement (faster, plausibly closeable inside the IPO window) or build a separate EU framework via the EBA and the EU AI Office (slower, more Brussels-friendly). The procedural difference is the difference between a May or June close and an October-or-later one. The Sunday-wire shape suggested a working-group structure rather than an immediate access decision, which would be a holding pattern that buys time for the federal-onramp executive action and any UK-bank access expansion to publish first. Three jurisdictions are now formally on the question: US (controlled-access via Glasswing, civilian expansion blocked), UK (AISI preview plus regulator-vendor working group, banks expansion reportedly within weeks), EU (no access today, formal Eurogroup discussion now in progress).
Code with Claude SF Two Days Out — Sonnet 4.8 Watch Window Now Inside May 6–13, Agentic-SDLC Track Is the Spine
The Code with Claude San Francisco developer conference lands Wednesday May 6, 8:00 AM to 8:00 PM PT, free livestream registration still open. The published agenda spine is the agentic-AI-in-the-SDLC track: how teams build production-grade agents on the Claude Platform, how teams run Claude Code at scale across long-horizon tasks, multi-repo work, parallel agents, and the infrastructure around them. Named speakers include Ami Vora (Head of Product) and Boris Cherny (Head of Claude Code). The Code with Claude: Extended track on Thursday May 7 is independent-developer and early-stage-founder focused: founder stories, builder deep-dives, hands-on workshops from the Applied AI team. London follows May 19, then Tokyo.
The Sonnet 4.8 window is now inside the conference week. Anthropic’s typical cadence puts the next Sonnet generation 1–4 weeks after the corresponding Opus release; Opus 4.7 shipped April 16, which puts the Sonnet 4.8 expected surface squarely inside the May 6–13 corridor. The leaked Claude Code source-map references from March named KAIROS persistent agents and Undercover Mode in the same package — both feature surfaces that would land alongside or just after Sonnet 4.8. Expected pricing: $3 / $15 per MTok, unchanged from Sonnet 4.6. If the Sonnet 4.8 ship date is announced at Code with Claude SF, that is the second major capability surface in a 10-day window after Claude Security, with the funding round mark and the Wall Street JV both already in place.
Claude Code Update Lands Windows / PowerShell 7 Detection, Drive-Letter Permission Anchoring, Pasted-OAuth Login Path
The latest Claude Code release continues the spring cleanup ahead of Code with Claude SF. PowerShell 7 installed via the Microsoft Store, MSI without PATH, or as a .NET global tool is now detected, and when the PowerShell tool is enabled Claude Code treats PowerShell as the primary shell instead of defaulting to Bash. Permission rules with drive-letter paths are now correctly root-anchored, and paths that differ only by drive-letter case are recognized as the same path — the longest-running Windows quirk in the changelog. Auto-mode UX gets a small fix: the spinner now turns red when a permission check stalls instead of looking like the tool is still running. The --dangerously-skip-permissions flag now bypasses prompts for writes to .claude/, .git/, .vscode/, shell config files, and other previously-protected paths, with catastrophic removal commands still gated as a safety net.
Auth login also picks up the WSL2 / SSH / container fix that has been in the issue tracker for months: claude auth login now accepts the OAuth code pasted into the terminal when the browser callback can’t reach localhost. The cumulative Windows-and-remote-dev posture is now the cleanest it has been since the Code product launched. Going into the conference, expect a feature-laden ship in the 48 hours after Wednesday’s keynote — the typical pattern. Production estates pinning for the conference week should hold the current build until at least Friday May 8.
Rate Limits API and Managed Agents Now Together Are the Operational Spine — What to Pin Before Conference Day
The Rate Limits API the Claude Developer Platform shipped on April 25 lets administrators programmatically query the rate limits configured for their organization and workspaces — the right primary instrument for the rolling-7-day cumulative outage budget conversation that emerged after last week’s three-incident Tuesday-Wednesday-Thursday cluster. Managed Agents, in public beta on the Claude Platform since April 8, gives teams a hosted harness with secure sandboxing, authentication, and tool execution handled for them — pricing is the standard Claude model usage plus eight cents per agent runtime hour, with web search billed at $10 per 1,000. Notion, Rakuten, Asana, Sentry, and Vibecode are the named early adopters; the integration paths for code automation, productivity, HR, and finance workflows are all live.
For shops planning conference-day load, the operational stack to pin: Managed Agents on the harness layer, Rate Limits API on the budget layer, flex tier on Bedrock plus secondary failover to Vertex on the inference layer. The Claude status page is clean going into Monday morning — five consecutive incident-free days for Claude.ai, the Anthropic API, Claude Code, and the Bedrock and Vertex tiers. No postmortem has yet shipped for the April 28 78-minute multi-surface event; the typical inside-ten-business-days cadence puts publication around May 8 to May 11, the same window as Code with Claude SF and Extended.
Claude Security Public Beta Updates Land — Repo, Directory, and Branch Scans Without Extra API Integration
Help Net Security and SC Media pushed updated coverage of Claude Security today. The public beta, powered by Opus 4.7, now scans entire repositories, specific directories, or individual branches without requiring additional API integration — a meaningful change for security teams that ran the private preview on stitched-together tooling. Each finding ships with a confidence score, severity, likely impact, reproduction steps, and a recommended patch. The product remains available to all Claude Enterprise customers globally; Team and Max tiers are next. The differentiator that survived the analyst circuit: Claude Security is positioned as the substrate the SAST and posture-management vendors integrate with, not a competitor to them — the bounded-disruption frame from CSIS and Forrester the week prior is the working narrative going into the IBM Think Boston track.
The Cyber Verification Program the Opus 4.7 release introduced is the gating mechanism for security professionals using the model for legitimate vulnerability research, penetration testing, and red-teaming. Watch for additional vendor-coalition releases through the IBM Think Boston window May 4–7 — CrowdStrike, Microsoft Security, Palo Alto, SentinelOne, TrendAI, and Wiz all have presence at the conference and overlap into the Code with Claude SF window. The two-track narrative — offensive Mythos plus defensive Claude Security — will get more public reps this week than at any point since the original April 7 Glasswing announcement.
Harvard FAS Phases Out ChatGPT Edu, Adds Claude — Anthropic’s Higher-Ed Foothold Widens Inside the May Calendar
The Harvard Crimson reported last week that the Faculty of Arts and Sciences will add Anthropic’s Claude to the suite of AI platforms available to affiliates while discontinuing access to ChatGPT Edu. The transition fits the pattern Anthropic has been running with art and design programs — the first three named curriculum partnerships are Art and Computation at Rhode Island School of Design, Fundamentals of AI for Creatives at Ringling College of Art and Design, and the MA/MFA Computational Arts program at Goldsmiths, University of London. The cumulative effect is Anthropic moving from research-lab brand into institutional stack at the front door of higher education, in parallel with the financial-services and creative-tools channel pushes already in motion.
The Higher-Ed track now overlaps with the IPO calendar in three ways: it gives the S-1 a clean education-vertical revenue line, it produces a steady stream of named-institution wins for the channel-momentum chart, and it builds Claude familiarity inside the cohort of students whose hiring patterns will set the next decade of enterprise procurement. The Q2 enterprise-vendor calls in May and June will likely pick up the Harvard FAS transition as a marker.
The Briefing: Financial Services Event Tomorrow May 5 — Anthropic Leadership and Institutional Deployments on the Same Stage
Anthropic’s livestreamed Briefing: Financial Services event lands tomorrow May 5. The published agenda has Anthropic leadership sharing upcoming developments and institutions deploying Claude at scale demonstrating what they have built. The timing is now load-bearing: the Wall Street JV announcement could plausibly fold into the same surface, the Eurogroup-Mythos readout from today will set the European regulatory frame the briefing has to address, and the financial-services analyst circuit will treat the event as the first formal post-blacklist statement of where Anthropic sees its non-federal regulated-industries footprint heading. This is also the first major event after the funding round news and the Pentagon contract awards landed.
Read for the channel: the briefing surface is where the SI partner mix — Accenture, BCG, Deloitte, Infosys, PwC — meets the financial-services-specific deployments at scale. If the Wall Street JV announcement does land at or alongside the briefing, the structural narrative becomes the cleanest it has been all year: Anthropic is not just an enterprise AI vendor selling API capacity, it is the substrate underneath three vertical channels (security, creative tools, financial services) with named global SI partners and named anchor investors on each. That is the S-1 narrative the IPO process needs.
The Pentagon-Plus-Wall-Street Split Is the New Anthropic Story: Federal Lockout, Private-Sector Acceleration, Mythos in Its Own Lane
Today’s news arrangement reads like the cleanest single-frame summary of Anthropic’s strategic position the year has produced. The Pentagon contract awards on May 1 lock in the federal procurement track without Anthropic on it. The $1.5B Wall Street JV nearing close demonstrates that the private-sector channel push is moving so fast the federal lockout is essentially a separate accounting line. And the Mythos national-security-moment carve-out shows that the only piece of the federal story that actually moves in the IPO timeline is the offensive-capability lane that the NSA is already running. Three tracks, three different velocities, three different counterparties — and the company’s “all lawful purposes” safety policy is what produced the split.
The week ahead concentrates this. Today: Eurogroup-Mythos discussion with Germany / France / Italy aligned on access parity. Tomorrow: The Briefing Financial Services event, plausibly the venue for the Wall Street JV announcement. Wednesday: Code with Claude SF, with the Sonnet 4.8 watch window inside the May 6–13 corridor. Thursday: Code with Claude Extended and IBM Think Boston wrapping the security-vendor conversation. The four-piece structure from yesterday compresses further: the EU regulator track opens this morning, the financial-services channel anchors tomorrow, the developer-and-capability track lands Wednesday, the security-vendor analyst frame closes the week. Every single one of these is being run while Anthropic is locked out of the federal procurement table. That is now a feature of the story, not a bug — the bull case is that the private-sector velocity makes the federal track optional inside the IPO window. The bear case is that an October S-1 with an active Pentagon lawsuit on the docket is a meaningfully harder pitch than the calendar without one.