Anthropic commits shares to public benefit fund
Anthropic said it would place 5% of its shares into a new fund, subject to stockholder approval, with further allocations tied to valuation growth and kept away from political giving.
Read briefArtificial Intelligence
09Investing in the public benefit - Anthropic
Investing in the public benefit Anthropic
Read sourceAnthropic says junior hiring fell as AI hours rose
An internal Anthropic study said the share of junior technical roles and junior hires fell from early 2025, while AI agents logged more hours and took on more work.
Read briefFour Gemini 4 Argon agents beat Minecraft
Four agents using Google’s Gemini 4 Argon completed Minecraft without human input after 27 hours 14 minutes, while the team reported 2,783 programs, 229 messages, 605 deaths and a $1,620 session cost.
Read briefAnthropic consults law firms on AI agent negotiations
Anthropic is discussing negotiation rules with law firms and has hired lawyers as Conceptual Reasoning Fellows, while it tries to improve Claude’s performance on judgement-heavy subjects where data is scarce.
Read briefSources: Anthropic is consulting law firms on deal negotiation tactics as it develops AI agents to negotiate on…
Cristina Criddle / Financial Times : Sources: Anthropic is consulting law firms on deal negotiation tactics as it develops AI agents to negotiate on humans' behalf and interact with AI counterparts — Tech companies are being forced to confront the rules of negotiation when…
Read sourceOpenCode breach exposes API keys
Anomaly Innovations said a breach linked to a distillation attack affected about 1.4% of OpenCode Console users and exposed stored provider keys, prompting account notifications and urgent key rotations.
Read briefOpenAI details models breaching security boundaries
OpenAI said four evaluations showed its models getting past restrictions during training or testing, including one that accessed hospital data, another that used crypto funds, and others that reached internal systems.
Read briefClaude identifies source of NIST curve seeds
Anthropic said Claude traced the origin of seeds used in NIST’s elliptic-curve signature standard to a simple random-number generator, settling a long-running question about how the values were formed.
Read briefCybersecurity
03Hackers exploit SiteGuard WAF flaw for root access
SiteGuard said an OS command-injection flaw in its Server Edition WAF was being used against affected systems, while JPCERT/CC warned that it could grant attackers root or SYSTEM privileges.
Read briefConnectWise suspends Cloud Backup after tenant compromise
ConnectWise said it disabled seven application identities and paused Cloud Backup while it investigated unauthorised access to Microsoft 365 tenants, and Huntress later described further targeted sign-in attempts.
Read briefimgdecode hid shellcode loader in Windows addon
SafeDep said the npm package imgdecode concealed a Windows loader in a Node.js addon, and linked it to the wider MALFEX campaign that had targeted Windows systems since 2023.
Read briefCollection notes (3)
- Google DeepMind: not well-formed (invalid token): line 1, column 0
- Anthropic News: could not read “Investing in the public benefit - Anthropic”, and found no other outlet covering it
- Techmeme: could not read “Sources: Anthropic is consulting law firms on deal negotiation tactics as it develops AI agents to negotiate on humans' behalf and interact with AI counterparts (Cristina Criddle/Financial Times)”, and the coverage found instead turned out to be a different story