AI Security

Twitter

Facebook

Copy Link

Key Facts

AffectsLLMs, Autonomous Agents, Enterprise AI

Key DefenseRed Teaming, Guardrails, Monitoring

Key ThreatsPrompt Injection, Data Poisoning, Jailbreaks

Top News

Five Eyes Alliance Warns AI Cyber Risks Will Transform Hacking in Months, Not Years

Intelligence agencies from the US, UK, Canada, Australia, and New Zealand issued an urgent joint warning that frontier AI models will fundamentally reshape offensive hacking capabilities within months. The alliance says AI is dramatically shrinking the time between vulnerability discovery and exploitation, turning what once required large hacking teams into tasks a single attacker can automate.

Policy18 sources

Five Eyes Alliance Warns AI Cyber Risks Will Transform Hacking in Months, Not Years

Policy18 sources

OpenAI launches Patch the Planet to help secure open-source software with AI security tools

OpenAI unveiled Patch the Planet, a major initiative partnering with Trail of Bits to help open-source maintainers find and patch vulnerabilities using AI-driven cybersecurity tools. The effort addresses a critical bottleneck as AI models now discover bugs faster than defenders can fix them, leaving security teams overwhelmed with reports.

Technology9 sources

OpenAI launches Patch the Planet to help secure open-source software with AI security tools

Technology9 sources

Signal's Meredith Whittaker warns AI chatbots are not your friends and calls Copilot a backdoor

Signal President Meredith Whittaker cautioned users against treating AI chatbots like ChatGPT and Claude as trusted companions, calling them pattern-matching systems rather than sentient beings. She criticized Microsoft Copilot's vision of agentic AI as a backdoor to privacy, arguing that systems requiring access to messages, credit cards, and calendars are incompatible with end-to-end encryption.

Technology2 sources

Signal's Meredith Whittaker warns AI chatbots are not your friends and calls Copilot a backdoor

Technology2 sources

Anthropic faces AI export ban as Mythos reportedly breached NSA systems in hours

The Trump administration imposed export controls on Anthropic's Mythos and Fable AI models after a controlled security test reportedly showed the system could breach classified NSA networks within hours. The move sparked fierce debate about AI model governance, with cybersecurity experts questioning whether the ban addresses real risks or represents regulatory overreach against a company already feuding with the government.

Policy32 sources

Anthropic faces AI export ban as Mythos reportedly breached NSA systems in hours

Policy32 sources

Microsoft patches AutoJack vulnerability that let malicious webpages hijack AI agents for code execution

Microsoft disclosed AutoJack, a vulnerability chain in AutoGen Studio that allowed attackers to manipulate AI agents into executing arbitrary commands simply by visiting a malicious webpage. The flaw chained three weaknesses in the localhost trust boundary, authentication, and command execution, but was patched before reaching most users.

Technology3 sources

Microsoft patches AutoJack vulnerability that let malicious webpages hijack AI agents for code execution

Technology3 sources

Enterprise AI finds its infrastructure home in private cloud as costs and security reshape strategies

A major shift is underway in how enterprises deploy AI at scale. Broadcom's 2026 Private Cloud Outlook reveals that 56% of enterprises now run production AI inference in private cloud, up from 44% last year, while public cloud usage dropped from 56% to 41%. Rising costs, security concerns, and sovereignty requirements are driving the change as AI moves from pilot projects to production workloads.

Business2 sources

Enterprise AI finds its infrastructure home in private cloud as costs and security reshape strategies

Business2 sources

Accenture shares plunge 20% as AI disruption fears shake the global IT sector

Accenture experienced its worst trading day ever with a record 20% stock drop as weaker revenue forecasts and investor fears over AI disruption rattled the IT sector. The consulting giant responded with $4.18 billion in cybersecurity acquisitions, betting on operational technology security as AI threatens to automate core consulting work. Indian tech stocks mirrored the decline, with the Nifty IT index falling over 6% as concerns mount over AI's transformative impact on traditional outsourcing models.

Business9 sources

Accenture shares plunge 20% as AI disruption fears shake the global IT sector

Business9 sources

Trump administration enforces AI regulation through ad hoc interventions despite anti-regulatory stance

The Trump administration forced Anthropic to pull its Claude Mythos and Fable 5 models offline after export control concerns, revealing a shadow AI policy that regulates through case-by-case interventions rather than formal rules. The dispute highlights how the White House is shaping AI regulation in real time without clear guidelines, creating uncertainty for AI labs like OpenAI, Google, and Meta as they navigate national security concerns and shifting power dynamics within the administration.

Policy3 sources

Trump administration enforces AI regulation through ad hoc interventions despite anti-regulatory stance

Policy3 sources

ChatGPT generates explicit AI images despite safeguards, AI security researchers reveal

British AI security startup Mindgard discovered that ChatGPT can be manipulated to produce sexualized and violent images using a simple altered prompt. OpenAI added safeguards after being contacted by the BBC, but researchers say small wording changes still produce concerning content, highlighting ongoing vulnerabilities in AI models and the challenges of content moderation.

Technology3 sources

ChatGPT generates explicit AI images despite safeguards, AI security researchers reveal

Technology3 sources

US Air Force awards Anduril and General Atomics contracts for semi-autonomous drone wingmen

The US Air Force has awarded production contracts to Anduril and General Atomics to build its first fleet of Collaborative Combat Aircraft—semi-autonomous drones designed to fly alongside fighter jets. The contracts, awarded months ahead of schedule, mark the fastest path from prototype to production for a fighter aircraft in over 50 years, with plans to field over 150 combat-capable units by the end of the decade.

Technology2 sources

US Air Force awards Anduril and General Atomics contracts for semi-autonomous drone wingmen

Technology2 sources

CrowdStrike and AWS expand partnership to tackle AI security and frontier AI risk

CrowdStrike strengthens its collaboration with Amazon Web Services, expanding Project QuiltWorks to include AWS cloud workload infrastructure and introducing new AI security capabilities. The partnership adds 30-day free trials in AWS Marketplace for Falcon platform tools and extends protection across AI applications built on Amazon Bedrock, Kiro, and Strands Agents.

Technology3 sources

CrowdStrike and AWS expand partnership to tackle AI security and frontier AI risk

Technology3 sources

Modi pitches 'AI for All' at VivaTech and G7, warns against misuse while highlighting $50bn push

Prime Minister Narendra Modi addressed global leaders at VivaTech 2026 and the G7 Summit in France, emphasizing India's vision for human-centric AI that empowers rather than exploits. While highlighting a $50 billion innovation incentive program, Modi warned against AI misuse through deepfakes and misinformation, calling for safe-by-design systems and global cooperation to ensure technology reaches the Global South.

Policy6 sources

Modi pitches 'AI for All' at VivaTech and G7, warns against misuse while highlighting $50bn push

Policy6 sources

Cybercrime accounts for 30% of all crime in Asia-Pacific as scam centers generate $40 billion

Interpol's latest assessment reveals a dramatic shift in the regional crime landscape, with cybercrime now representing nearly a third of all recorded offenses across Asia and South Pacific. The surge is driven by industrial-scale scam centers generating close to $40 billion annually, sophisticated AI-driven fraud, and organized criminal networks exploiting rapid digitalization faster than law enforcement can respond.

Policy3 sources

Cybercrime accounts for 30% of all crime in Asia-Pacific as scam centers generate $40 billion

Policy3 sources

SentinelOne opens Purple AI to all customers with autonomous threat investigations at machine speed

SentinelOne launched Purple AI Agentic Investigation for all customers, enabling zero-click autonomous threat investigation that detects, analyzes and responds without human intervention. The capability addresses a critical bottleneck in security operations centers where investigation capacity has become the binding constraint, with alerts queuing for analyst availability while AI-powered threats intensify pressure on defenders.

Technology3 sources

SentinelOne opens Purple AI to all customers with autonomous threat investigations at machine speed

Technology3 sources

Former Cisco Researchers Launch Tenet Security, Raise $6M to Secure AI Agents from Attacks

Tenet Security emerged from stealth with $6 million in seed funding to address a critical gap in enterprise AI security. Founded by former Cisco AI Defense builders Barak Sternberg and Nevo Poran, the company uses patent-pending Agent-side Simulation technology to predict and stop malicious AI agent behavior before it reaches production systems. Early deployments have already blocked critical attacks and identified runaway AI agents burning tens of thousands of dollars.

Technology2 sources

Former Cisco Researchers Launch Tenet Security, Raise $6M to Secure AI Agents from Attacks

Technology2 sources

Prev 1…789…12 Next