Safety & Security

AIAI-generated summary Malicious actors launched fake ChatGPT and Claude applications on GitHub to distribute Yoba malware while bypassing detection. The campaign uses signed executables and DLL sideloading to deploy a cryptocurrency miner on compromised Windows computers.

AIAI-generated summary Google Threat Intelligence Group and Mandiant researchers have available findings showing that criminals break into corporate cloud accounts to run AI models at the victim's expense. Mandiant states that some attacks resulted in bills reaching up to $1 million before the attackers lost access.

AIAI-generated summary CrowdStrike and Deloitte announced an expanded cybersecurity partnership in Barcelona to connect business risk decisions with threat detection and response. Deloitte will integrate Falcon IQ into its advisory services and use NVIDIA open Nemotron models to build security plans for clients.

AIAI-generated summary OpenAI rolls out higher security bug bounties that raise top payouts to $172,000 to encourage independent researchers to find vulnerabilities. OpenAI says it will apply the new reward ranges to previously submitted reports that have not yet received an award.

AIAI-generated summary According to reports, security experts are warning that AI agent swarms are targeting infrastructure and raising concerns about power grid security. Governments around the world are telling people to prepare for at least 72 hours of disruption due to these potential threats.

AIAI-generated summary According to reports, a fanfiction clone site scraped Archive of Our Own and charged for audiobook versions before shutting down after receiving over 2,000 takedown requests. The archive repository Archive of Our Own remains offline following a security incident while authors express concern over unauthorized scraping and monetization of their works.

AIAI-generated summary Security researchers have released details about a technique known as memory poisoning, which allows attackers to inject malicious or misleading memory items into an artificial intelligence assistant's persistent memory layer. Forcepoint states that the aim of these attacks is to trick artificial intelligence agents into producing false information, creating a persistent risk that can repeatedly output flawed data across future sessions.

AIAI-generated summary HackerOne has released the 2026 HackerOne Security Research Report featuring figures on AI-assisted submissions and researcher payouts. The research shows that 85 percent of respondents are actively building their AI skills while valid AI vulnerability reports rose 222 percent.

AIAI-generated summary The Crown Prosecution Service released a statement after a senior prosecutor cited an imaginary legal precedent in a high court case. The Crown Prosecution Service states it has a strict policy that limits artificial intelligence use and requires human checks for accuracy.

AIAI-generated summary Glow Labs releases research documenting AI coding agents installing deprecated servers from archived repositories. The installation of these unpatched packages creates high-severity security vulnerabilities including remote code execution risks on developer machines.

AIAI-generated summary OpenAI has released updates to its Security Bug Bounty program featuring higher rewards and new incentives for independent researchers finding software vulnerabilities. OpenAI says the changes are designed to encourage researchers and build a collaborative security environment while recognizing findings that help fix issues.

AIAI-generated summary OpenAI announced plans for its next internal security sprint, partnering with Cathedral Systems to test defenses and quadrupling rewards for critical findings in its Security Bug Bounty program. OpenAI says the initiative builds on previous efforts where engineers used Codex to investigate findings and prepare patches as part of an ongoing effort to strengthen systems.

AIAI-generated summary University of Washington researchers released a new study showing that AI models from Google, OpenAI and Alibaba penalize speakers with Nigerian and Chinese accents during interviews and tests. The study indicates that the models generally ranked American and British speakers highest while giving lower scores to Chinese and Nigerian speakers based on their delivery and perceived clarity.

AIAI-generated summary Anthropic announced in a regulatory filing that it may imminently meet its internal definition for recursive self-improvement capability. Anthropic says using artificial intelligence to help build more capable systems could pose considerable safety risks if progress sharply accelerates.

AIAI-generated summary Cloudflare has made available new controls that let website owners distinguish between search crawlers, AI training bots, and agents acting on behalf of users. The new settings default to allowing search crawlers while blocking AI training bots and agents on pages displaying ads, though website owners can adjust those rules.

AIAI-generated summary Meta launched its Muse AI agent in September to run on its own cloud computer so no other agent can access it. Financial institutions are facing a wave of customer-authorized bot activity and growing security concerns regarding third-party tools.

AIAI-generated summary Google rolled out a host of Gemini-enabled features across Gmail, including voice-activated commands and live AI search functions. Google states that Gemini does not train its foundational AI models on personal emails and only accesses inboxes for isolated tasks.

AIAI-generated summary Anthropic is rolling out text watermarking more broadly to include older systems such as Mythos 5, Opus 4.7, Opus 4.6, and Sonnet 4.6. The watermarking system applies globally and cannot be opted out of, with only eligible organisations able to detect the watermark under EU law.

AIAI-generated summary Cisco made available a security harness specification called Foundry designed to act as a governance layer for large language models. Cisco says the implementation helps teams turn model output into a reliable process by verifying citations against bytes on disk.

AIAI-generated summary VicOne released an initial set of 28 zero-day findings identified in less than two days using AI-assisted research across software supporting the Physical AI ecosystem. The discoveries span three stages of the robotics lifecycle and create demands for robotics teams to assess how each weakness affects robot safety and operations.