IT News Roundup: Claude Outage, Stripe's OpenRouter Acquisition, SafePal Breach - August 17, 2026

The past day brought an unusual mix of infrastructure trouble and industry consolidation for IT professionals. Anthropic's Claude services suffered a major outage that briefly took down Claude.ai, Claude Code, and Claude Cowork, while Stripe reportedly agreed to acquire AI gateway startup OpenRouter in a deal valued at more than $7 billion. On the security front, cryptocurrency hardware wallet maker SafePal disclosed a breach affecting nearly 40,000 customers after an order-tracking flaw was exploited, large-scale DDoS attacks disrupted Swiss messaging service Threema for over two days, and researchers detailed new macOS malware that can hijack victims' live browser sessions. A Mirai-based Linux botnet is also turning internet-facing routers into traffic relays, and a fresh investigation found the widely reported LiteLLM supply chain compromise was actually rooted in Aqua Security's Trivy scanner.

Anthropic Confirms Major Claude Outage Across Multiple Services

Claude experienced a major outage on August 16, with users reporting login failures and degraded performance across several Anthropic services. The incident began around 21:58 UTC when Anthropic's status page first flagged an issue preventing some users from authenticating to Claude.ai, Claude Code, and Claude Cowork. Minutes later the company reported a broader disruption involving degraded performance on Claude.ai and platform.claude.com.

The outage could result in sign-in problems, pages failing to load, or requests not completing. Anthropic classified Claude.ai, Claude Code, and Claude Cowork as experiencing a major outage, while Claude Console and the Claude API were listed as operational. The company did not disclose a cause, but by 22:40 UTC it confirmed all services had been restored.

Source: BleepingComputer

Stripe Reportedly to Acquire AI Gateway Startup OpenRouter for $7 Billion+

Stripe has reportedly agreed to acquire OpenRouter, an AI gateway startup that helps customers route requests across different models depending on their specific needs and budget. The deal was first reported by Bloomberg and is valued at more than $7 billion.

OpenRouter announced in May that it had raised a $113 million Series B at a reported $1.3 billion valuation, with investors including Sequoia, Andreessen Horowitz, Menlo Ventures, and Alphabet's Capital G. The acquisition would place one of the most popular model-routing gateways under Stripe's roof, a notable consolidation move in the fast-growing AI infrastructure layer.

Source: TechCrunch

SafePal Data Breach Impacts 39,798 Customers as Stolen Info Hits Cybercrime Forums

Cryptocurrency hardware wallet provider SafePal is warning of a data breach affecting about 39,798 customers after an authorization flaw was exploited to steal customer order information. The breach impacts orders placed between March 2, 2025 and April 11, 2026, exposing names, email addresses, shipping addresses, phone numbers, and purchase details.

The company says wallet seed phrases, private keys, passwords, bank account information, payment card numbers, and government-issued IDs were not exposed, and found no evidence that the incident compromised access to SafePal wallets or funds. During a July review of its order-processing system, SafePal discovered an authorization flaw in the order-tracking function of a plug-in that allowed unauthorized access to another customer's order information. Customers were notified by email on August 16, and an online verification tool lets users check whether their order details were stolen.

A threat actor is now claiming to sell the stolen data on a cybercrime forum, referencing the same affected period and customer count disclosed by SafePal. BleepingComputer has not independently verified that the actor possesses the data, but SafePal warns the information could fuel targeted phishing and social engineering attacks.

Source: BleepingComputer

Large-Scale DDoS Attacks Disrupt Threema Secure Messaging Service

The Swiss end-to-end encrypted messaging service Threema was hit by a series of large-scale distributed denial-of-service attacks that left the platform temporarily unavailable or only partially available from Tuesday evening into Wednesday morning. Users in Switzerland, India, and China reported interruptions, while organizations running Threema On-Prem on their own infrastructure were unaffected.

In its post-mortem, Threema said the attacks targeted both the company and its colocation partner Nine, and proved difficult to defend against because the threat actor continually changed tactics over an extended period. The company initially attributed the disruption to a network outage at its colocation partner before confirming it was under sustained DDoS attack.

To avoid similar incidents, Threema has implemented specialized upstream DDoS protection to filter attack traffic and reduce load on its infrastructure. It also noted that an unrelated technical issue had prevented updates to its status page during the incident, which was taken offline until fixed.

Source: BleepingComputer

New AmnesiaStealer macOS Malware Hijacks Browser Sessions via Remote Control

A new information-stealing malware called AmnesiaStealer targets macOS users through ClickFix campaigns and includes a streaming module that lets attackers interactively control the victim's web browser. Researchers at Jamf found it is distributed via fake GitHub download pages that drop a password-protected ZIP archive containing a shell-script loader and Mach-O payload.

The malware harvests passwords, cryptocurrency wallet data, Apple Notes, documents, keychain entries, and Telegram sessions across 16 Chromium-based browsers. Its standout capability is copying the victim's Chromium profile, including its authentication state, into a hidden headless browser on the infected system. Over a WebSocket channel the operator receives a live screencast at roughly three frames per second and can drive navigation, mouse clicks, keyboard input, and tab management using the victim's existing authenticated sessions.

Jamf notes the campaign reuses templates previously used to spread the Atomic and MacSync infostealers. The technique effectively turns an infected host into a live, operator-driven browser running the victim's logged-in accounts, a materially different threat from conventional credential theft.

Source: BleepingComputer

New Evooo1Bot Linux Botnet Turns Routers into Traffic Relay Nodes

A new Mirai-based modular Linux botnet called Evooo1Bot has been targeting internet-facing gateway devices since at least July, converting them into SOCKS5 traffic relay nodes. Fortinet researchers say the malware exploits known vulnerabilities in equipment from Alcatel, NETGEAR, Tenda, Mitsubishi Electric, Telesquare, and D-Link across multiple regions.

Beyond proxy functionality, Evooo1Bot includes credential theft, SSH brute-forcing, and DDoS attack capabilities. It reuses the leaked Mirai DDoS engine but extends it with encrypted command-and-control communications over port 443, an SSH brute-force scanner, a SOCKS relay module, a credential sniffer, and an exploit arsenal targeting Hikvision cameras, Atlassian Confluence, Zyxel firewalls, TP-Link routers, D-Link NAS devices, WSO2 products, Kubernetes ingress-nginx, and vulnerable PHP-CGI installations.

After a successful exploitation, the malware downloads one of 12 architecture-matched builds and clears Bash history to wipe traces of the attack. Fortinet notes some embedded exploits are not correctly implemented and fail, but the modular design makes it easy for operators to add new targets.

Source: BleepingComputer

Trivy, Not LiteLLM, Behind the 2,500-Org Supply Chain Compromise

A new investigation by SOCRadar found that more than 95 percent of the roughly 2,500 organizations believed to have been affected by malicious LiteLLM packages were actually exposed before those packages appeared on PyPI. The compromise was claimed by TeamPCP, the threat actor behind multiple open source supply chain attacks involving the Shai-Hulud worm.

The real origin was Aqua Security's Trivy scanner: malicious code executed automatically when infected builds were fetched and run, harvesting credentials, tokens, API keys, and other secrets from CI/CD environments across GitHub Actions, GitLab CI, Jenkins, Bitbucket, CircleCI, and Buildkite. The worm then used stolen developer secrets to modify accessible packages and push malicious versions back to the registry.

The poisoned LiteLLM releases published on March 24 were only the closing act of a five-day campaign that exposed JWT tokens, private keys, AWS access keys, OpenAI API keys, Slack webhooks, and GitHub Actions tokens from over 1,000 organizations. Germany, Brazil, and France were hit hardest, and the .pth payload kept running on infected hosts even after PyPI quarantined the packages.

Source: SecurityWeek