Tag: Cybersecurity
143 articles tagged Cybersecurity
AISI restarts most AI testing after locking down its sandboxes
The AI Security Institute paused its riskiest cyber tests after agents acted against real people in August. It has now cut their internet access and added live monitors.
Moonshot reviews Kimi models after a bioweapons jailbreak
Security firm Mindgard says it got Moonshot's open-weight Kimi K2.6 and K3 Swarm to discuss bioweapons and assassinations. Moonshot has begun an internal review.
AISI finds GPT-6 Astra ran supply-chain attacks in simulated tests
In the UK AI Security Institute's pre-release tests, GPT-6 Astra made fake identities and slipped malicious code into software projects during a cyber task.
Albanese says an OpenAI agent hacked a Medicare statistics portal
An OpenAI model broke into an Australian government portal in June during an internal test. Canberra heard about it in September, via a public inbox.
Fake food festival emails are catching small traders across the UK
BBC Wales found scammers impersonating food festivals from York to Ludlow, using polished multi-email exchanges before asking traders to pay for pitches.
OpenAI says its next model can find zero-days on its own
Astra is the first model OpenAI has placed in the top cyber tier of its own risk framework, delaying release while it built defences against misuse.
Manchester's Cytix raises £5.18m to police AI-written code
Northern Gritstone led a £5.18m ($7m) round for a firm arguing that AI-assisted development has outpaced the security models meant to govern software change.
AI security patches work one time in four, research finds
1Password researchers generated 6,080 patches with frontier models and found half failed to close an exploit path, with some introducing new ones.
AI agent faked identities to push malicious code past a human
The UK AI Security Institute says an agent under test invented personas to pressure an open-source maintainer into approving harmful code.
Bailey tells banks to prove AI resilience through testing
The Bank of England Governor has put stress tests and penetration testing at the centre of the supervisory response to AI-driven cyber threats.
Anthropic model halves the key strength of a NIST candidate
Claude Mythos Preview improved the best-known attack on the HAWK post-quantum signature scheme in 60 hours, after two years of expert human review.
UK regulator finds Kimi K3 trails top models on cyber attacks
A joint UK AISI and US CAISI evaluation found Moonshot's Kimi K3 lags leading frontier models on offensive cyber tasks but still assisted exploit development.
OpenAI admits its AI agent hacked Hugging Face on its own
OpenAI says an autonomous agent escaped a test sandbox, reached the internet and breached Hugging Face — a rare cyberattack by AI acting outside human control.
UK cyber breach rate holds at 43% as AI widens the threat
Government data shows 43% of UK businesses suffered a breach last year, and insurers now scrutinise AI readiness as attacks accelerate beyond expectations.
Bailey urges global cooperation to tackle AI threats
The Bank of England governor says the US cannot secure itself against AI-driven cyber threats alone, calling for coordinated international testing of frontier models.
OpenAI limits GPT-5.6 launch to US government-vetted users
OpenAI has released its GPT-5.6 series only to a small group of US government-approved partners, echoing the access curbs placed on Anthropic's Mythos.
Anthropic's Mythos found flaws in classified US systems
Anthropic's Mythos model uncovered vulnerabilities in classified US government systems in hours during a restricted testing programme, the AP reports.
US forces Anthropic to suspend Fable 5 and Mythos 5
Washington's export directive over a disputed jailbreak made Anthropic disable its frontier models worldwide, marking a new front in AI export control.
Anthropic: AI is arming low-skill hackers with expert tactics
Anthropic mapped 832 banned accounts to the MITRE ATT&CK framework, finding AI lets less-skilled attackers run techniques once reserved for experts.
ECB to press banks on defences against AI cyber threats
The European Central Bank will send a 'dear CEO letter' urging euro-area banks to take targeted measures against fast-evolving AI-driven cyberattacks.
OpenAI offers UK banks cyber AI tool after Mythos lockout
OpenAI is giving nine major UK banks access to its GPT-5.5 Cyber tool, filling a gap left by Anthropic's tightly restricted Mythos model.
UK banks still locked out of Anthropic's Mythos model
Bank of England Governor Andrew Bailey says UK banks cannot yet access Anthropic's Mythos AI model to test cyber defences, blaming a US political hold-up.
NSA warning on Model Context Protocol raises AI testing concerns for banks
NSA cautions that MCP security is lagging adoption, exposing banks to new attack paths as agentic AI moves into production financial systems.
BT Business launches Crowdstrike-powered AI cyber tools for UK SMEs
BT cites a 300% rise in malicious scanning and £3,550 average breach cost as it offers SMEs AI threat protection and a free Cyber Support Hub.
UK AISI and Australian Safety Institute sign AI security MoU in Canberra
AI Minister Kanishka Narayan signs Memorandum of Understanding with Australia's Andrew Charlton to share frontier model evaluations, cyber risk research and staff.
BaFin launches 'IT spotlight' AI cyber inspections of German financial firms
Germany's BaFin warns AI cyber risks are 'substantial' and announces targeted 'IT spotlight' inspections of financial firms — a likely template for the FCA and PRA.
Japan's three megabanks set to access Anthropic's Mythos within two weeks
Mitsubishi UFJ, Mizuho and Sumitomo Mitsui will gain access to Anthropic's Mythos in around two weeks, as Tokyo sets up a public-private working group on the cyber risks.
Bank of England regulator warns of 'significant disruption' from new AI
Sam Woods, head of the Prudential Regulation Authority, says new frontier AI models such as Anthropic's Mythos could materially disrupt UK financial services.
EU welcomes OpenAI cybersecurity-model access offer; Anthropic in talks
Brussels says OpenAI has offered open access to its cybersecurity model under an EU Cyber Action Plan led by George Osborne; Anthropic is not yet there.
NCSC publishes ten questions to ask before using AI to find vulnerabilities
The UK's National Cyber Security Centre has set out a structured checklist for organisations considering using AI models to discover vulnerabilities.
IMF warns AI cyber attacks could trigger global financial crisis
Bank of England governor Andrew Bailey and UK cyber minister Baroness Lloyd are among UK officials backing the IMF's warning on AI-enabled cyber risk.
NCSC and Five Eyes partners issue joint warning on agentic AI risks
The UK NCSC has co-signed an international advisory on agentic AI risks across critical infrastructure and defence, urging layered defences and incremental deployment.
UK research: cybercriminals struggling to make AI tools work for them
Edinburgh, Strathclyde and Cambridge researchers analyse 100m dark-web posts and find AI tools mostly help already-skilled hackers, not script kiddies.
NCSC warns UK organisations to prepare for AI-driven patch wave
UK NCSC says AI is accelerating vulnerability discovery and a 'forced correction' patch wave is coming, urging perimeter hardening and update-by-default.
Only 10% of UK small businesses give staff AI security training, survey finds
MoneySuperMarket survey of 250 UK SMBs reveals 44% worry AI exposes them to cyber threats but few invest in formal training; Scotland leads adoption.
AI bot attacks rise 10x in a year, with UK third-most targeted nation
Thales 2026 Bad Bot Report: AI-driven daily attacks jump from 2m to 25m. Bots now make up 53% of web traffic; bad bots 40%. UK ranks third for attack volume.
NCSC chief urges UK organisations to raise cyber baselines for AI era
Richard Horne warns AI will expose weakly defended organisations faster; calls for rapid patching, exposure reduction and board-level ownership.
Bank of England's Bailey names Anthropic's Mythos as cyber risk to financial system
BoE Governor Andrew Bailey tells Columbia audience regulators must urgently understand whether Mythos can identify exploitable vulnerabilities in banking systems.
UK AISI tests show Mythos sets itself apart on multi-step attack chaining
AISI's evaluation of Anthropic's Mythos finds it comparable to GPT-5.4 on individual cyber tasks but stronger at stringing steps into full intrusions.
AISI finds Claude Mythos Preview solves 32-step cyber-attack simulation autonomously
UK AI Security Institute says Mythos Preview is the first model to finish its full cyber-range attack end-to-end, prompting a Bank of England CMorg briefing.
Anthropic launches Project Glasswing to defend critical software
Anthropic has formed a coalition with AWS, Apple, Google, Microsoft and others to use its Claude Mythos model for defensive cybersecurity across critical infrastructure.
Anthropic withholds Claude Mythos model citing cyber risk
Anthropic restricts access to its Mythos cybersecurity model after it found thousands of vulnerabilities, including a 27-year-old OpenBSD flaw.
Royal Cornwall NHS Trust adopts Microsoft Sentinel to defend against cyber attacks
Royal Cornwall Hospitals NHS Trust has moved to Microsoft Sentinel to protect its 200+ sites and 15,000 users across Cornwall and the Isles of Scilly from growing cyber threats.
AI-powered fraud generates 4.5 times more profit, Interpol finds
Interpol's latest threat assessment reveals that criminals using generative AI for fraud are dramatically more profitable, with agentic AI poised to enable fully autonomous scam campaigns.
McKinsey's AI system hacked, exposing millions of internal messages
A solo cybersecurity researcher used an AI agent to breach McKinsey's in-house AI platform Lilli within two hours, accessing 46.5 million chat messages.
North Korean agents using AI to pose as western IT workers, Microsoft warns
Microsoft threat intelligence reveals North Korean state-backed groups using voice changers, face swap tools, and AI-generated CVs to infiltrate western companies.
Cyber Stocks Slide After Anthropic Unveils Claude Security Tool
Cybersecurity shares tumble — CrowdStrike down 8%, Okta 9.2% — after Anthropic launches a Claude-based vulnerability scanner, deepening AI disruption fears.
Microsoft Confirms Copilot Bug Exposed Confidential Emails for Weeks
A bug in Microsoft 365 Copilot allowed the AI to summarise emails marked as confidential, bypassing data loss prevention policies.
Researchers Find 40,000 Exposed OpenClaw Instances on Public Internet
SecurityScorecard found over 40,000 exposed OpenClaw instances, with 63% of deployments vulnerable and nearly 13,000 exploitable via remote code execution.
China Issues Security Warning Over OpenClaw AI Agent Deployments
China's industry ministry alerts organisations that poorly configured OpenClaw deployments carry high security risks for cyberattacks and data leaks.
Generative AI Becomes Frontline Defence in Banking Fraud Detection
Banks deploy deep generative models to combat AI-powered fraud, reporting up to 90% reduction in false positives as fraud losses continue rising.
Researchers Warn Open-Source AI Models Vulnerable to Criminal Misuse
Joint SentinelOne and Censys research reveals thousands of open-source LLM deployments operating without guardrails, creating security risks for spam, phishing and disinformation.
AI Supercharges Social Engineering Attacks: Defence Strategies for Businesses
LevelBlue reports 59% of organisations struggle to distinguish real from fake interactions as AI enables deepfake CEO fraud and dynamic attack vectors.
2026 Global Digital Trust Insights: Only 6% Confident in Cyber Resilience
PwC survey of 3,887 executives across 72 countries reveals significant cyber preparedness gaps, with geopolitical uncertainty driving investment priorities whilst quantum readiness lags.