AI Security

France’s blood service says alarms caught altered freezer settings at three sites. Plasma was moved, with no reported product loss or health impact.

Binance says a fake recruitment process, malware and access to an internal AI tool helped attackers reach its wallet signing system.

Google says it disrupted the XCPE botnet, which used malware embedded in TV-box firmware. The backdoor may remain on affected devices.

AtomicChat says its DeepSeek V4.1-Flash edit cut safety-test refusals to zero. The result is unverified and can run offline on a 12 GB GPU.

Retailers in at least six countries report refrigeration system attacks. The method resembles an October incident in France; AI involvement is unconfirmed.

Cisco Talos says VOLTA sends profiles built from stolen Windows data to AI models for target assessments. The report does not confirm a model generated one.

Cisco Talos says VOLTA sends stolen victim data to AI models for assessment. Anthropic is also extending an invisible text watermark to older Claude models.

October brought record CVE disclosures, rising exploitation and major crypto losses. AI helped find and fix flaws, but the evidence is mixed.

TRM Labs says most of roughly $6.36 billion left Binance wallets in about 20 minutes. The episode renews debate over AI-driven security and its risks.

Berlin’s ransomware breach exposed about 1.4 million files. A broadcaster found the detection system covered only 5% of devices in one department.