Anthropic accuses Alibaba of model theft

The big story today is Anthropic going public with allegations that Alibaba-linked operators extracted data from Claude via nearly 29 million fraudulent exchanges. It's a rare direct accusation between major AI players, and it lands alongside a Cisco zero-day that went undetected for two months and a Klue breach that's hitting security vendors themselves.

Sources

AI theft allegations

The Anthropic-Alibaba story is the day's most consequential piece, with implications for AI security and US-China tech tensions.

#01www.bbc.comJun 25

Anthropic accuses Chinese rival Alibaba of illicitly extracting AI capabilities

Anthropic claims Alibaba-linked operators made nearly 29 million fraudulent exchanges with Claude to extract capabilities. This is a rare direct accusation between major AI firms and could escalate into a larger geopolitical flashpoint.

5 min readRead original

Zero-days and breaches

Two incidents show attackers are getting stealthier and more targeted, hitting both network infrastructure and data platforms.

#02undercodetesting.comJun 25

Cisco SD-WAN Zero-Day CVE-2026-20245: The Root-Level Backdoor That Lurked For Two Months + Video - Undercode

Mandiant found that attackers exploited a critical Cisco SD-WAN zero-day (CVE-2026-20245) as early as March, two months before Cisco disclosed it. The flaw gives root access to authenticated attackers, a reminder that zero-days can lurk longer than vendors admit.

#03www.theregister.comJun 25

Security shops among the 'hundreds' of Klue hack victims

The Klue data breach, perpetrated by the Icarus extortion crew, is hitting multiple cybersecurity companies including Huntress. It's a supply-chain attack on a market intelligence platform, and the victim list is still growing.