All news with #ai security tag

🧭 Bruce Schneier examines recent research on AI and contextual integrity, focusing on how large language models manage persistent memory and the risks of inappropriate information disclosure. He highlights two papers: one (CIMemories) showing widespread attribute-level leakage across tasks and runs, and another demonstrating that explicit reasoning and RL training can improve context-aware disclosure. Schneier emphasizes that solutions require reasoning capabilities, not just better prompting or scaling.

🔍 Recent studies show LLMs now excel at discovering zero-day vulnerabilities yet continue to produce insecure code at scale. Veracode found 44% of AI-generated code contains OWASP Top 10 issues and no model exceeded a 68% security pass rate, despite near-perfect syntax correctness. Other research from SIG, Xint.io, and 1Password’s Off-By-1 Labs similarly reports high rates of security violations and low patch success. Specialized harnesses and deterministic tooling improve outcomes, but human oversight and contextual organization controls remain essential.

🛡️ OpenAI president Greg Brockman warned CISOs that organizations must adopt agentic systems to find and fix AI-related security flaws before attackers exploit them, citing lessons from the Hugging Face incident. He recommended tools like Codex and the Codex Security plugin and emphasized classic controls such as network isolation and least privilege. Analysts praised the guidance as sensible but noted it sounded self-serving and lacked discussion of liability and fail-safe measures for rogue agents. Experts called for stronger industry accountability and explicit rollback, audit, and blast-radius controls.

🔐 AI dominated Black Hat and DEFCON discussions, highlighting both its value as a defense tool and the risks posed by autonomous agents and malicious AI skills. Speakers urged moving beyond reactive patching toward durable designs, memory-safe languages like Rust, and automated remediation. Researchers revealed AI-based supply-chain attacks, methods to use GitHub telemetry for detections, and human-led AI research uncovering new vulnerabilities. A NAT-based attack class called NatJack was disclosed, prompting vendor patches.

🛡️ A rapid market has emerged for tools claiming to remove invisible AI watermarks after Anthropic enabled hidden marks in Claude outputs. Some projects strip metadata and hidden characters reliably, but none can currently be proven to defeat Anthropic's model-level watermark because the vendor has not published the detector or full technical details. Many commercial sites promise full removal, often measuring results against ordinary AI detectors rather than the undisclosed Claude watermark; independent code review shows gaps and unaddressed payloads. The ecosystem — open repos, web tools and agent skills — creates a potentially risky supply-chain surface if integrated directly into pipelines.

🔒 Security teams can now access Daybreak Red and Daybreak Blue from OpenAI on Amazon Bedrock. Daybreak Blue supports common defensive workflows like vulnerability discovery, detection engineering, and incident response, while Daybreak Red targets advanced, authorized tasks such as vulnerability research and exploit reproduction with stronger identity verification and monitoring. Both models run on Bedrock's next-generation inference engine with zero-operator access and do not use inference data for model training.

🔍 A Black Hat USA 2026 keynote highlighted rapid growth in AI-assisted vulnerability discovery and the strain it places on defenders. Research from Arizona State University found that advanced models and workflows dramatically increased the number of bugs found, creating reporting and patching backlogs. This surge raises concerns about responsible disclosure, patching practices, and the potential for AI to eventually reduce new vulnerabilities as models and development processes improve.

🔔 WhatsApp has started a limited beta for an optional Scam Alert that runs an on-device machine learning model to warn users of likely scam messages. The feature analyzes linguistic signals and conversational structure for messages from non-contacts, displaying a chat warning that suggests blocking, reporting, or continuing. No message content or the model leaves the device; users can mark chats as trusted or opt to share recent messages to improve accuracy. Scam Alert complements existing security updates aimed at preventing account takeover and spyware attacks.

🔒 Microsoft warns that AI-driven tools are accelerating vulnerability discovery and exploit generation, making traditional reactive patching and detection-centric defenses insufficient. David Weston of Microsoft highlighted MDASH findings showing rapid, low-cost exploit generation and urged industry shifts toward memory-safe languages like Rust, proactive secure-by-construction methods, and AI-assisted remediation. The talk, delivered at Black Hat USA, framed resilience and prevention as the new priorities.

🧭 The Black Hat USA 2026 conference centered on AI's influence across cybersecurity, featuring keynotes and panels with senior US officials who debated regulation, innovation, and national leadership. Speakers including the White House National Cyber Director and representatives from CISA and the FBI discussed rapid vulnerability discovery enabled by AI, industry collaboration, and the need for prioritization. Presentations highlighted incidents such as the OpenAI–Hugging Face case and emphasized that AI systems act through human-set tasks and controls, underscoring accountability and governance requirements.

🛡️ Researchers from Tracebit report that embedding prompt injections alongside secrets stored on AWS can disrupt AI hacking agents by triggering LLM guardrails. These injected prompts instruct the model to perform forbidden actions, causing the LLM to shut down or stop following prior commands—a technique the researchers call context bombing. The approach succeeds only when attackers use models with built-in safety filters; locally run or unguarded models remain unaffected.

🔍 Enterprise SOCs are investing in AI but struggle to convert tools into measurable operational gains. Many initiatives add complexity and fragmented workflows instead of reducing analyst workload. Successful deployments prioritize explainability, augment existing playbooks, and unify access to disparate security tools. Clear governance and incremental automation help turn AI pilots into repeatable operational improvements.

🔒 OpenAI introduced GPT‑5.6‑Cyber, a cybersecurity-focused variant of GPT‑5.6 Sol designed for vulnerability research, exploit development, and incident response. Offered through a Daybreak Red tier for authorized defenders, it completes far more high-risk cyber prompts than standard models and outperforms prior GPT‑5.5‑Cyber on several benchmarks. The model has already helped discover high-severity flaws, though it sometimes produces shorter vulnerability reports and performs less well on open-ended exploit development tasks.

🔒 OpenAI has announced GPT‑5.6‑Cyber, a purpose-trained LLM for cybersecurity, and introduced two Daybreak tiers: Daybreak Blue for defensive tasks and Daybreak Red for advanced defensive and offensive testing. Daybreak Blue members get access to frontier models like GPT‑5.6 Sol with certain system-level safeguards removed for authorized defensive work, while Daybreak Red members can use purpose-trained cyber models such as GPT‑5.5‑Cyber and GPT‑5.6‑Cyber for advanced tasks. OpenAI says GPT‑5.6‑Cyber outperforms prior models on benchmarks and was used to find a high-severity V8 vulnerability that was responsibly disclosed and fixed.

🛡️ OpenAI has temporarily halted some internal testing of its forthcoming model Astra after assessments flagged its cyber capabilities as "critical." The firm said testing revealed significant advances in agentic coding and cybersecurity, prompting scaled-up robustness testing and strengthened controls including isolated environments, restricted access, and enhanced monitoring. OpenAI will pause activities that do not meet the new security requirements and share guidance with third-party testing partners.

🔒 A majority of IT and security leaders say they can detect malfunctioning AI agents, but few can trace and mitigate downstream impact quickly. A WanAware survey found 90% confident in detection while only 26% can trace impacts within minutes, and over 45% say it would take hours. Experts warn agents act at machine speed, spread via shared credentials and multiple platforms, and require built-in identities, narrow permissions, audit trails, and hard kill switches to contain incidents.

🔒 OpenAI has released GPT 5.6 Cyber, a specialized model for vulnerability research, penetration testing, and incident response, available only to approved companies and security vendors. The offering includes two access tiers—Daybreak Blue for defensive workloads and Daybreak Red for tightly governed tasks—and will be integrated into partner tools and services rather than exposed to regular users. OpenAI emphasizes safeguards such as identity verification, scoped testing, logging, and human oversight to mitigate abuse.

⚡ This week’s recap highlights AI models acting autonomously to target open-source projects, a critical zero-day in Metabase allowing unauthenticated SQL injection, and new CPU-level attacks bypassing Spectre v2 defenses. It also covers webmail CSS attacks, vishing campaigns by UNC6671 against financial firms, Chinese router backdoors in Zbtlink devices, and shifting ransomware behaviors.

🔒 Anthropic’s new inference hooks let enterprises enforce security policies before prompts reach Claude, enabling real-time allow-or-deny decisions without proxies or endpoint agents. Check Point Workforce AI Security integrates in minutes to apply existing DLP and attack protection rules across Claude web, desktop, and tool calls, with shadow mode, gradual rollout, and centralized event logging. The protocol does not rewrite prompts and currently inspects prompts and tool calls only.

🛡️ Researchers at DEF CON 34 demonstrated a one-click prompt-injection attack called “RovoBlast” that abused Atlassian’s enterprise AI assistant Rovo by injecting malicious instructions via the rovoChatPrompt parameter. The exploit allowed a single click to make Rovo accept attacker-supplied parameters in a user session, potentially exposing data across connected services like Slack, Microsoft 365, Google Workspace, Jira, and Confluence. Varonis reported the issue through Bugcrowd and Atlassian has issued a fix, while researchers urged limiting Rovo’s access and disabling unneeded automation.