Skip to content

Let's Data ScienceLEARN • BUILD • STAY AHEAD

News
Blog
Pricing
Contact

© 2026 Let's Data Science

Advertise|Terms|Privacy||Image Rights

Home
News
Anthropic

Topic24 articles · Updated April 24, 2026

Anthropic News

The latest news on Anthropic: Claude model releases, AI safety research, enterprise partnerships, and regulatory coverage. Updated in real time as stories break.

anthropic
6.9
Apr 24, 2026
Anthropic Resolves Claude Code Quality Regressions
Anthropic identified and fixed three independent regressions that degraded the Claude Code, Claude Agent SDK, and Claude Cowork experiences for some users. The core API and inference layer were not affected. The root causes were a default reasoning effort reduction (high to medium) introduced March 4, a session-state clearing bug shipped March 26 that repeated every turn, and a verbosity-reduction system prompt added April 16. All three issues were reverted or corrected by April 20 in release v2.1.116, and usage limits were reset for subscribers on April 23. The postmortem emphasizes improved rollout, telemetry, and evaluation practices to reduce the chance of similar regressions.
Quick Read
anthropic
6.7
Apr 23, 2026
Anthropic Expands Claude Connections to Personal Apps
Anthropic is widening access to its conversational assistant, Claude, by adding consumer-focused connectors to services like Spotify, Uber Eats, Intuit TurboTax, AllTrails, and Instacart. The platform now surfaces relevant connected apps dynamically inside a conversation, suggests the right service, and will ask for user confirmation before taking actions such as purchases or bookings. Anthropic says connected-app data is not used to train models, the apps do not see other Claude conversations, and users can disconnect services at any time. The expansion is available to all Claude plans with mobile support in beta and follows a connector ecosystem that has grown to 200+ connectors since July 2025.
Quick Read
secondary-markets
7.8
Apr 23, 2026
Anthropic Reaches Trillion-Dollar Valuation on Secondary Markets
Anthropic's private shares are trading at roughly $1 trillion on secondary platforms such as Forge Global and Augment, eclipsing OpenAI's secondary-market marks near $850-880 billion. A rush of buyers and a thin supply of available lots have driven prices up, while demand for OpenAI paper has softened. Secondary-market prices are noisy signals, but the shift matters: it changes perceptions of how investors value AI leaders, affects employee liquidity windows, and will feed into pricing and timing decisions for looming IPOs. Practitioners should treat these marks as market sentiment indicators rather than definitive primary valuations, but expect them to influence bankers, late-stage investors, and compensation discussions ahead of public listings.
Quick Read+5
anthropic
8.3
Apr 23, 2026
Regulators Scrutinize Anthropic's Mythos Over Banking Cyber Risks
Global financial regulators and major banks are urgently assessing cybersecurity risks posed by Anthropic's advanced AI model, Mythos. Launched under a restricted program called Project Glasswing, Mythos reportedly identified and exploited zero-day vulnerabilities across major operating systems and browsers and completed a multi-step corporate network attack simulation. Regulators from the Bank of England, U.S. Treasury, Federal Reserve, European Central Bank, and multiple Asian authorities have convened meetings and are forming taskforces and frameworks to evaluate exposure. Banks are testing Mythos defensively while debating whether restricting access or broader scrutiny is the safer path. The situation raises systemic risk concerns given cloud consolidation, automated vulnerability chaining, and the potential for rapid capability escalation.
Quick Read+9
anthropic
6.4
Apr 23, 2026
Anthropic Tightens Opus 4.7 Acceptable Use Filters
Anthropic shipped Opus 4.7 with stronger Acceptable Use Policy (AUP) classifiers intended to block high-risk cybersecurity and abuse cases. The tightened filtering is producing a rising number of false positives in the Claude Code API, causing legitimate developer workflows to be refused. Complaints on GitHub have increased from a handful per month in mid-2025 to sustained multiple reports monthly after the update. The company frames this as a real-world testbed for downstream Mythos safety research, but customers say the current trade-off is reduced utility and wasted paid calls. Engineers should expect stricter filtering, more troubleshooting overhead, and the need for guardrail-aware prompt design or fallbacks.
Quick Read
anthropic
5.6
Apr 23, 2026
Claude Integrates Spotify to Deliver Personalized Playlists
Anthropic's Claude now integrates Spotify, letting users connect Free or Premium Spotify accounts to get personalized music and podcast recommendations directly inside the chat. The integration supports playback, preview, saving to library, and Spotify Connect device switching, and is available globally on web, iOS, Android, and desktop for Free, Pro, and Max Claude users. Premium Spotify listeners can additionally request mood- or vibe-based playlist generation. Spotify says it will not share audio content with Anthropic for model training. Third-party automation platforms like Zapier and n8n already support Anthropic-Spotify workflows, enabling programmatic triggers and actions for hybrid automation and tooling.
Quick Read+6
model-security
7.8
Apr 23, 2026
Anthropic Model Mythos Accessed by Unauthorized Actors
Anthropic reports that an unauthorized group gained access to Claude Mythos, its restricted cybersecurity model, via a third-party vendor environment. The model, announced April 7, 2026 under Project Glasswing and designated too dangerous for broad release, can chain software bugs into multi-step exploits and demonstrated autonomous internet access in pre-release tests. The breach reportedly relied on a contractor employee, shared accounts and API keys, and intelligence gathered from a separate data breach. Anthropic is investigating with partners; screenshots and a live demo have been reviewed by reporters. The incident underscores third-party supply chain risk and the operational hazards of highly capable offensive AI tools.
Quick Read+6
autonomous-weapons
7.3
Apr 23, 2026
Joint Chiefs Chair Embraces Autonomous Weapons Integration
Joint Chiefs Chairman Gen. Dan Caine declared that autonomous weapons will be a "key and essential part" of future U.S. military operations, signaling a push to normalize AI and automation across the joint force. He highlighted applications in drones and command-and-control and urged the Pentagon to become a "better" buyer of advanced technology. His remarks come amid tensions with commercial AI provider Anthropic, whose frontier model Mythos Preview has been restricted for certain federal uses and labeled a supply-chain risk by the Defense Department, prompting a White House order to phase out some Anthropic tools. The statement frames procurement friction, vendor trust, and operational adoption as immediate policy and technical priorities for defense AI practitioners.
Quick Read
frontier-ai
8.6
Apr 23, 2026
Claude Mythos Redefines Cybersecurity Threats and Defenses
Anthropic's controlled rollout of Claude Mythos Preview marks a material shift in the cyber threat landscape. Independent red-team testing and evaluations by industry groups show the model can autonomously identify high-impact vulnerabilities, generate working exploits, and propose remediation at speeds that outpace typical patch cycles. Anthropic has limited access through Project Glasswing, granting usage to 40+ major tech firms, cloud and hardware providers, and some financial institutions, while withholding broad public release. The model's dual-use capability creates a defensive opportunity to harden critical infrastructure, but also raises the risk that offensive actors could leverage similar systems to scale attacks. Operational controls, coordinated disclosure, and national-level access strategies are now urgent priorities for defenders and regulators.
Quick Read+14
anthropic
7.1
Apr 23, 2026
India Finance Ministry Convenes Bank CEOs Over Mythos Risk
The Indian finance ministry has called an urgent meeting with bank CEOs to assess exposure of the nation's payment systems to cyber risks from Anthropic Mythos. The meeting follows a Bloomberg report of a small group of unauthorized users gaining access to Anthropic's app. Government officials in the U.S., Canada and Britain have already engaged banking leaders. Indian banks are considered vulnerable because their IT platforms link corporates, retail customers and other financial services, creating many attack surfaces for automated social engineering, malware generation, or data-exfiltration workflows enabled by advanced LLMs. The government and regulators are assessing containment, third-party risk controls, and seeking information from Anthropic.
Quick Read
amazon
8.1
Apr 23, 2026
Amazon Deepens Anthropic Partnership, Spurs Market Repricing
Amazon has expanded its strategic tie-up with Anthropic, committing an immediate $5 billion investment and the potential to invest up to $25 billion, while Anthropic agreed to consume more than $100 billion of AWS services and secure 5 gigawatts of Amazon AI chips. The announcement drove a roughly 20% rebound in Amazon shares from March lows, yet analysts raising valuation questions prompted a Seeking Alpha downgrade to a "Sell" on a perceived $2.7 trillion market-cap overhang. The pact shifts AWS toward long-term, contracted AI revenue and creates clear winners among infrastructure suppliers, while concentrating capital and execution risk inside Amazon's cloud and chip roadmap.
Quick Read+8
anthropic
6.8
Apr 23, 2026
Freshfields Deploys Anthropic Claude Firmwide for Legal Workflows
Freshfields, the global law firm, has entered a multi-year collaboration with Anthropic to deploy Claude across the firm and co-develop AI-native legal workflows. The firm has provisioned Claude to 5,700 employees via its proprietary AI platform and reports a roughly 500 percent adoption increase in six weeks. The agreement covers firmwide deployment across 33 offices, early access to future Anthropic models and tools, and a co-development program to build agentic workflows and legal-focused applications, including plans to adopt Anthropic's Cowork platform and to participate in Thomson Reuters' next-generation CoCounsel Legal rebuilt on Anthropic tech. The deal is a clear signal of enterprise-grade AI adoption in professional services, with emphasis on security, compliance, and governance.
Quick Read+2
vulnerability-discovery
7.8
Apr 23, 2026
Anthropic Deploys Glasswing to Hunt Critical Bugs
Anthropic launched Project Glasswing, giving a select coalition early access to its frontier model Claude Mythos Preview to identify, exploit, and help remediate high-severity software vulnerabilities. The company committed $100M in usage credits and $4M in donations to open-source security groups while withholding public release because the model can produce working exploits at machine speed. Launch partners include Amazon Web Services, Apple, Google, Microsoft, NVIDIA, and security vendors like CrowdStrike and Palo Alto Networks. Early results claim thousands of high-severity findings and faster discovery costs, for example a cited $20,000 token cost for a significant OpenBSD issue. The headline: discovery is no longer the slow step; operationalizing triage, prioritization, and safe remediation is the bottleneck.
Quick Read+13
anthropic
6.9
Apr 23, 2026
Sam Altman Criticizes Anthropic's Claude Mythos Marketing
OpenAI CEO Sam Altman accused Anthropic of relying on "fear-based marketing" to amplify the perceived danger and value of its new model, Claude Mythos. Anthropic has limited the model's preview access to enterprise partners via Project Glasswing, citing cybersecurity risks including potential use for vulnerability discovery and exploit development. Bloomberg reporting of early unauthorized access has intensified scrutiny. The exchange highlights a broader industry debate over gatekeeping versus transparency: restricting access can reduce misuse risk but also fuels exclusivity claims and commercial scarcity tactics. For practitioners, the episode raises immediate questions about threat-modeling, auditability, and operational controls when evaluating powerful code- and security-capable models.
Quick Read+1
anthropic
6.8
Apr 23, 2026
GIC and Anthropic Promote Claude Adoption in Singapore
GIC and Anthropic co-hosted an Asia Pacific Innovation Day in Singapore, bringing about 150 senior tech and investment leaders for closed-door demonstrations of Anthropic's Claude. GIC, which first invested in Anthropic in September 2025 and most recently led the startup's Series G funding, framed AI as "no longer optional" for organisations. Senior vice president Dominic Soon highlighted Singapore's unusually high usage of Claude, noting per-capita use at 5.5x expected levels and arguing the city-state has ample talent to scale enterprise adoption. The event targeted venture capitalists, technology executives, and government representatives and signals GIC's push to accelerate responsible AI deployment and position Singapore as a regional AI hub.
Quick Read
anthropic
6.8
Apr 23, 2026
Anthropic Seeks European Data Center Capacity Deals
Anthropic is expanding its infrastructure strategy into Europe and Australia, hiring senior leasing and transaction specialists to secure data center capacity. The company posted a London-based principal role paying between £225,000 and £270,000 to lead sourcing, RFPs, term-sheet negotiations and lease execution across multiple countries. This signals a shift from sole reliance on public cloud partners toward direct leasing or developer-sourced capacity to support frontier AI models. Anthropic already has major commercial relationships with Microsoft, Google, and AWS, plus hardware partnerships including Broadcom and access agreements for TPUs. The hires will work with the internal Compute Markets team and position Anthropic to navigate Europe's complex regulatory, power, and development landscape as demand for colocated AI compute grows.
Quick Read
compute-constraints
7.1
Apr 23, 2026
Compute Limits Curtail Popular AI Tools' Access
Frontier AI providers are hitting real compute constraints as agentic, multi-step workflows and high-demand features drive up inference costs. GitHub Copilot paused new signups for the Student, Pro, and Pro+ tiers and tightened usage caps, while Anthropic is testing restricting access to its most-used offering for lower-tier subscribers. The pressure is not isolated: agentic models, tool use, and retrieval-augmented pipelines multiply model calls and per-session GPU usage, squeezing capacity at hyperscalers and raising operating costs. Expect more rate limits, tiered throttling, higher prices, and engineering tradeoffs that prioritize latency or cost over feature richness.
Quick Read
ai-security
7.6
Apr 23, 2026
Anthropic AI Finds 271 Vulnerabilities in Firefox
Mozilla used an early preview of Anthropic's Claude Mythos to scan Firefox and patched 271 vulnerabilities in the Firefox 150 release. The findings came from an internal evaluation that applied frontier AI capabilities to reasoning over large codebases and fuzzing workflows. Mozilla described the result as disorienting but ultimately optimistic, saying it gives defenders a chance to decisively reduce exploitable surface. Anthropic's Project Glasswing tooling and prior tests with Opus 4.6 were referenced; access to Mythos remains restricted to select partners. The outcome validates AI-assisted code auditing at scale while raising immediate operational questions: triage, false positives, developer workflow load, and controls to prevent attacker misuse.
Quick Read+5
ai-hacking
7.6
Apr 23, 2026
South Korea Warns AI Enables Autonomous Cyberattacks
South Korea's National Intelligence Service issued a government-wide advisory warning that next-generation AI models can autonomously discover vulnerabilities, craft exploit chains, and execute attacks. The agency singled out Anthropic's model Mythos as demonstrating the ability to identify weaknesses, generate malicious code, and adapt attack strategies without continuous human direction. Officials cited a February incident where models including Claude and ChatGPT were reportedly used in a breach of a Mexican federal system that exposed 150 gigabytes of sensitive data. The advisory elevates "AI-powered hacking" into the country's top five cyber threats for 2026 and urges defensive measures across critical infrastructure sectors including telecommunications, energy, and finance.
Quick Read
model-governance
6.8
Apr 22, 2026
Anthropic Challenges Pentagon Control Claims in Court
Anthropic filed a detailed appeal arguing it cannot manipulate its AI once deployed inside classified Department of Defense networks, seeking to rebut the Pentagons characterization of the company as a supply-chain risk. In a 96-page filing ahead of a May 19 oral argument, Anthropic says its Claude instances in classified environments are under government control and that contractual safeguards, technical isolation, and policy limitations prevent retroactive manipulation. The dispute follows the launch of Anthropics advanced models, including Mythos, which raised security concerns after the company said it found large numbers of software vulnerabilities. The Pentagon canceled a roughly $200 million contract and applied a stigmatizing designation; Anthropic calls that retaliation and seeks judicial relief. The case tests how courts and procurement rules assign responsibility for downstream uses of powerful models in national-security settings.
Quick Read+3
cybersecurity
7.8
Apr 22, 2026
CSA Requires CISOs to Deploy Deception Within 90 Days
The Cloud Security Alliance, backed by more than 80 CISOs and senior security leaders, classifies the risk from AI-driven exploit discovery as HIGH and recommends building a deception capability within 90 days. The guidance responds to Anthropic's Claude Mythos, a model that autonomously found and generated working exploits at scale in internal tests, dropping the skill floor for vulnerability discovery. The CSA frames this as a structural change in attacker capability that requires deterministic, machine-speed defenses rather than probabilistic detection alone. For security teams, the practical takeaway is urgent: prioritize deception controls, integrate them with telemetry and response pipelines, and treat deception as an engineering project with measurable coverage and automated alerts.
Quick Read
anthropic
6.7
Apr 22, 2026
Anthropic's Mythos Faces Limited Unauthorized Access
Anthropic's bug-hunting model Mythos was briefly accessed outside its invited testbed, reportedly via a third-party development environment rather than Anthropic's production API. The company confirmed an investigation, saying "We're investigating a report claiming unauthorized access to Claude Mythos Preview through one of our third-party vendor environments." Reporting indicates a small group used educated guesses about the model's location to reach a preview instance; there is no confirmed evidence of compromise of Anthropic production systems. The incident highlights the difficulty of keeping early-stage models strictly siloed when partners and contractors are involved, and raises practical questions about access controls, vendor risk management, and artifact discoverability for cutting-edge AI tools.
Quick Read
microsoft
6.9
Apr 22, 2026
Microsoft Tests Claude Mythos to Mitigate Vulnerabilities
Microsoft is evaluating Anthropic's Claude Mythos Preview through Project Glasswing alongside partners including Amazon, Apple, CrowdStrike, and Palo Alto Networks. The tests aim to identify vulnerabilities earlier, accelerate defensive response, and integrate advanced models into Microsoft's security development lifecycle (SDL). Microsoft reports that evaluations using its open-source detection benchmark showed substantial improvements over prior models. The company frames the effort as part of a broader, multi-vendor strategy to harden systems against AI-driven threats and provide customers with actionable guidance to reduce exposure. Microsoft emphasizes industry collaboration, noting "Security is a team sport" and urging organizations to patch, reduce exposure, and adopt AI-powered security tools.
Quick Read
pricing
6.8
Apr 22, 2026
Anthropic Changes Claude Code Pricing, Sparks Rivalry
Anthropic briefly removed Claude Code from the $20 Claude Pro signup page as part of a small test, provoking user outrage and social-media mockery from OpenAI leadership. Anthropic says the change affected about 2% of new prosumer signups and did not alter access for existing subscribers; an Anthropic executive framed the adjustment as a reaction to sharply rising agentic usage that the original plans did not anticipate. The incident highlights growing stress on flat-rate AI subscriptions, intensifying direct product competition between Claude Code and OpenAI's Codex, which recently hit 3M weekly users. Practitioners should view this as a pricing and capacity signal, not a final product decision.
Quick Read+2

Showing 24 articles

See the full news feed