Skip to content

Let's Data ScienceLEARN • BUILD • STAY AHEAD

News
Blog
Code Problems
Pricing
Contact

© 2026 Let's Data Science

Advertise|Terms|Privacy||Image Rights

Home
News
Cybersecurity

Topic24 articles · Updated June 23, 2026

AI Cybersecurity News

The intersection of AI and cybersecurity: model vulnerabilities and CVEs, prompt injection and jailbreaks, adversarial attacks on ML systems, and the defensive tooling being built to push back.

five-eyes
7.2
Jun 23, 2026
Five Eyes Urges AI to Counter Cyber Threats
In a rare joint statement, Five Eyes cybersecurity agencies from Australia, the United States, the United Kingdom, Canada and New Zealand warned leaders must "act swiftly" as artificial intelligence increases the "speed, scale and sophistication of cyber threats," the UK NCSC and ABC News report. The agencies argue that AI also offers powerful defensive tools and list urgent steps for organisations, including reassessing internet-exposed systems, reducing unnecessary connectivity, fixing legacy software, tightening access controls, and elevating cyber resilience to board-level responsibility (UK NCSC; ABC News). Media outlets also note that on June 13 Anthropic suspended worldwide access to its two most powerful models, Fable 5 and Mythos 5, amid export-control and security concerns (SMH).
Quick Read+16
openai
7.1
Jun 23, 2026
IBM integrates OpenAI cyber AI into security
IBM has joined the OpenAI Daybreak Cyber Partner Program and launched a new application security service that applies OpenAI's cyber capabilities to enterprise code review and vulnerability validation, according to IBM's press release. The offering is delivered via IBM Consulting Advantage and operates inside client environments with read-only access to repositories and bounded execution. Project Lightwell, backed by a $5 billion commitment from IBM and Red Hat, will use OpenAI's cyber capabilities alongside other frontier models for code review and remediation. IBM's press release quotes Mark Hughes, Global Managing Partner for Cybersecurity Services, IBM Consulting, saying the Daybreak program expands IBM's access to advanced AI capabilities to surface risks faster.
Quick Read+2
agentic-ai
4.5
Jun 23, 2026
Enterprises Drive Agentic AI, Cloud and Cybersecurity Training
Economic Times CIO reports on edForce's FY2025-26 enterprise learning data, showing Agentic AI training grew 100% year-on-year to 57,536 learners. The same dataset shows cloud training up 29% to 41,876 learners, cybersecurity up 39% to 25,448 learners, and DevOps up 26% to 31,211 learners, per the report. Certification-linked learning grew 36% to 38,562 certifications. Ravi Kaklasaria, Co-Founder and CEO of edForce, is quoted saying enterprise learning is increasingly outcome-driven. These figures are per edForce's own data as reported by ET CIO and have not been independently audited.
Quick Read
five-eyes
7.9
Jun 23, 2026
Five Eyes Warns AI Could Threaten Governments, Businesses
Cybersecurity agencies from the Five Eyes alliance issued a rare joint statement - signed by NSA Cybersecurity Director David Imbordino and acting CISA Director Nick Andersen - warning that frontier AI models will transform offensive cyber capabilities in months, not years. The statement explicitly names Anthropic's Fable 5 and OpenAI's Daybreak, states 'the timeline is not years, it is months,' and urges organizations to address legacy systems, slow patching, and weak identity controls before AI-assisted attacks become routine. The Guardian notes the warning follows the Trump administration's export-control directive restricting foreign nationals from accessing Fable 5. CISA published the joint statement on its website, with all five member nations as co-signatories. The agencies recommend integrating cybersecurity into core business strategy rather than treating it as a compliance issue.
Quick Read+4
openai
7.6
Jun 22, 2026
OpenAI launches Patch the Planet to fix open-source bugs
OpenAI launched 'Patch the Planet' on June 22, 2026, a Daybreak initiative built with Trail of Bits, HackerOne, and Calif to find and fix vulnerabilities in widely used open-source software using GPT-5.5-Cyber and expert human review. Early field results include 8 Linux kernel pointer-leak proof-of-concepts, 24 local privilege escalation exploits, over 10 exploitable Safari vulnerabilities, a Firefox WebAssembly CVE (CVE-2026-8390) patched before Pwn2Own, and four of six dnsmasq CVEs independently flagged before their public fix. Initial projects include cURL, Python, the Go project, aiohttp, Sigstore, and pyca/cryptography. Separately, Anthropic's Project Glasswing using Claude Mythos has uncovered more than 10,000 high-or-critical vulnerabilities across the most systemically important software, with that program now expanded to roughly 150 organizations in over 15 countries.
Quick Read+4
anthropic
8.5
Jun 22, 2026
Mythos Allegedly Breaches Almost All NSA Classified Systems
The Economist, as cited by TechTimes, reports that Senator Mark Warner relayed that Gen. Joshua Rudd told him Anthropic's Mythos, in a June 11 authorized red-team exercise, "broke into almost all of our classified systems, not in weeks, but in hours." The Economist editor Shashank Joshi, who wrote the original quote, later clarified the claim "should not be read literally," noting it depended on Mythos working alongside other tools under specific conditions. On June 12, the U.S. Commerce Department issued an export-control directive barring all foreign nationals including Anthropic's own non-citizen employees from accessing Fable 5 and Mythos 5. Anthropic, unable to selectively comply, suspended both models for all users worldwide, per Fortune. Anthropic called the directive "a misunderstanding" and says the cited trigger was a narrow jailbreak - asking the model to read a codebase and fix flaws - that also applies to rival models including OpenAI's GPT-5.5. BeInCrypto and Digg document expert pushback; independent forensic verification of the red-team scope remains outstanding.
Quick Read+12
european-banking-authority
6.8
Jun 19, 2026
EBA Warns AI Models Increase Cyber Risks for Banks
The European Supervisory Authorities (EBA, EIOPA and ESMA) published their first annual overview of major ICT-related incidents under the Digital Operational Resilience Act (DORA) on June 3, 2026. The report, covering 3,383 major incidents from EU financial entities, states that banks must strengthen cybersecurity to keep pace with the potential use of highly capable AI-driven tools. The ESAs flag that one third of incidents had cross-border impact, driven by system failures and external events, and underline that robust third-party risk management is essential. The report signals that EU banking supervisors now formally treat advanced AI tools as a material cyber risk factor requiring operational resilience planning.
Quick Read
anthropic
8.6
Jun 18, 2026
US Directive Suspends Anthropic's Mythos and Fable Access
According to Anthropic's June 12 statement, the US government issued an export control directive that "suspend[s] all access to Fable 5 and Mythos 5 by any foreign national," and Anthropic said it disabled access to both models for all customers to comply. Anthropic reported receiving the letter at 5:21pm (ET) on June 12 and said the government did not provide detailed explanation of its national security concern. The Wall Street Journal reports that researcher Nicholas Carlini demonstrated how Mythos could be used to find critical software vulnerabilities; TechPolicy.Press reports that concerns raised to the White House, including by Amazon CEO Andy Jassy, helped prompt the government's action. CNBC and Fortune report Anthropic disputed that the demonstrated bypass constituted a universal jailbreak and characterized the government's action as lacking transparency, technical criteria, and due process. Anthropic said access to all other models was unaffected.
Quick Read+7
sovereign-ai
7.3
Jun 18, 2026
Dream raises $260M for sovereign AI infrastructure
PR Newswire and The Jerusalem Post report that Dream raised $260 million in a financing round that brings the company to a $3 billion valuation, three years after founding. The round was co-led by Bicycle Capital and Group 11, with participation from Antler, Bain Capital Ventures, Tru Arrow Partners, and others, according to PR Newswire. The company has said it has secured nearly $300 million in total contract value since commercial operations began in late 2024, per PR Newswire and The Jerusalem Post. Reporting by Globes and TNW notes the founders include former NSO Group CEO Shalev Hulio and former Austrian chancellor Sebastian Kurz, and Globes reports Dream's annual recurring revenue recently crossed $100 million. Editorial analysis: This round underscores investor interest in sovereign-AI and national cyberdefense offerings as governments weigh control and resilience of critical AI infrastructure.
Quick Read+5
llmjacking
6.5
Jun 18, 2026
Hackers Turn IT Tools Into Remote-Access Payloads
According to GBHackers reporting based on Sysdig telemetry, a financially motivated threat actor used a publicly exposed Ollama model server as the reasoning engine for an automated, multi-stage offensive framework. GBHackers says the attacker integrated unauthenticated model inference into a VAPT-style pipeline that performed service fingerprinting, vulnerability matching, web reconnaissance, payload synthesis, and command execution, with the model returning machine-parseable JSON at each stage. Sysdig captured the framework's architecture and prompt-engineered logic because the actor sent full-stage instructions on every inference call, GBHackers reports. The researchers labeled the tool "VAPT" and identified a reliable detection marker sequence used as an RCE check, echo VAPTb3gin; id; echo VAPTfin, GBHackers says. GBHackers adds that probe targets during the capture window were private practice ranges and lab networks (RFC 1918 and HackTheBox spaces), not external victims.
Quick Read
ai-agents
8.1
Jun 17, 2026
Low-skilled Attacker Uses Claude and Codex to Breach Firms
According to OALABS research published June 16, investigators analyzed over 1,000 agent sessions recovered from a compromised Linux staging host, documenting a single attacker using locally-installed Claude (claude-opus-4-6) and OpenAI Codex (gpt-5.2-codex) to breach at least 14 companies. Per OALABS, the attacker -- identified from session logs as residing in Addis Ababa, Ethiopia after an OPSEC failure -- directed agents via vague natural-language prompts framed as redteam exercises, automating reconnaissance, N-day exploit development (including CVE-2025-5777/CitrixBleed 2 and CVE-2025-54068/Livewire), credential harvesting, and data exfiltration. A Lightning Network node holding approximately 69.71 BTC was breached; wallet cracking was attempted but failed. Anthropic's November 2025 blog, cited for broader context, documents a separate state-sponsored AI-assisted espionage campaign.
Quick Read+2
ai-agents
5.5
Jun 17, 2026
Small Business Owners Deploy AI Teams to Run Operations
Black Enterprise reports that small business owners are increasingly managing teams of AI agents to run core operations including customer service, scheduling, bookkeeping, email and research. The outlet summarises a New York Times feature (June 4, 2026) documenting entrepreneurs using agent frameworks such as OpenClaw to execute multi-step workflows across applications and websites, and an Entrepreneur article (June 5, 2026) by Sherin Shibu reporting the same shift. Scott Bell, a bankruptcy lawyer, told the New York Times: "Of course, that's going to put me out of a job, as well, at some point." Black Enterprise also cites a Nasdaq Economic Institute report, referenced by Axios, finding one-person business formations have surged since early 2025, and quotes security researchers warning that autonomous agents raise cybersecurity and quality-control risks requiring human oversight.
Quick Read
databricks
7.0
Jun 16, 2026
Databricks acquires Panther Labs to bolster security
Databricks announced the acquisition of cybersecurity startup Panther Labs, according to SiliconANGLE. The companies did not disclose financial terms, and SiliconANGLE notes Panther had a $1.4 billion valuation after a 2021 funding round. SiliconANGLE describes Panther's platform as a telemetry-first detection and enrichment system that supports customer-provided detections, a natural-language detection authoring workflow, automated false-positive root-cause analysis, and a query syntax called PantherFlow. Reporting from intellectia.ai frames the deal as part of a broader Databricks push into security and cites commentary by CEO Ali Ghodsi about AI-driven threats. Databricks earlier introduced Lakewatch, an open, agentic SIEM, per CyberRiskLeaders.
Quick Read+1
cybersecurity
6.8
Jun 16, 2026
AI Enables Surge in Cyberattacks Against Internet Assets
The Atlantic reports that AI tools are increasing the scale, speed, and sophistication of cyberattacks. According to The Atlantic, the cybersecurity firm Palo Alto Networks identified a fourfold increase in daily attacks from 2024 to 2025 among its tens of thousands of clients, and the article quotes Alex Stamos saying, "There's a crazy amount of offensive activity happening right now." The Atlantic describes AI-written phishing, adaptive malware, automated espionage, and minute-scale data theft as contributing factors. Editorial analysis: For practitioners, the reported trends widen the threat surface and raise demands for higher-fidelity telemetry and faster incident response automation.
Quick Read+1
softbank
7.2
Jun 16, 2026
SoftBank Launches OpenAI Patching Service for Cybersecurity
Per reporting from The Associated Press and the Washington Post, SoftBank Group Corp. is launching a service that uses OpenAI technology to detect and "patch" cybersecurity weaknesses for large Japanese organisations. Masayoshi Son described the vulnerability to cyberattacks as "a crisis" and said the offering is "a patching service" targeting the nation's top 3,000 companies, including critical infrastructure operators, The Washington Post reports. The companies set up a 50:50 joint venture, SB OAI Japan, last year to develop and exclusively market AI services for Japan, according to the coverage. OpenAI CEO Sam Altman did not attend the Tokyo launch because his daughter was born earlier than expected, and Mark Chen, OpenAI's chief researcher, attended in his place, the reporting says. No financial terms were disclosed; SoftBank said attendees at the event can apply for a free diagnosis, per the coverage.
Quick Read+3
anthropic
7.2
Jun 15, 2026
Cybersecurity Leaders Ask US to Lift Anthropic Restrictions
More than 120 cybersecurity and technology professionals signed an open letter urging the Trump administration to lift export-control restrictions on Anthropic's Fable 5 and Mythos 5 models, according to PYMNTS. The letter, led by former Facebook Chief Security Officer Alex Stamos, was dated June 14 and addressed to Commerce Secretary Howard Lutnick and National Cyber Director Sean Cairncross, per PYMNTS and Axios. Reuters reports that signatories include security executives and researchers at Adobe, Zoom, Sophos, and Nvidia. The letter followed a US government directive that, per Anthropic's official statement, required the company to disable Fable 5 and Mythos 5 for all customers to comply with export controls barring access to any foreign national. Anthropic said the government cited a narrow jailbreak -- asking the model to read a codebase and fix software flaws -- a capability the company says is also available in other publicly deployed models.
Quick Read+2
phishing
7.6
Jun 15, 2026
Google Sues Cybercrime Network Over Gemini-Powered Phishing
The complaint filed in Manhattan federal court alleges a China-linked cybercrime network, identified by Google as the Outsider phishing kit, used AI tools including Gemini to automate large-scale phishing campaigns, according to Reuters and TechCrunch. Google's filing and media reporting link the operation to roughly 9,000 fake websites, one million fraudulent web domains, and 2.5 million text messages sent to Android users in a two-week span, per TechCrunch, and to more than 1.5 million URLs detected between November and April, per Reuters. The complaint asks the court to block the software and seeks monetary relief, Reuters reports. Reuters and TechCrunch say Google is coordinating with the FBI and U.S. telecom carriers AT&T, T-Mobile, and Verizon, and that law enforcement seized domains and storefronts in a joint action.
Quick Read+12
banking
4.5
Jun 14, 2026
Banks Confront AI Errors, Oversight and Security
According to Forbes, John Werner reports that AI in finance delivers speed and insight while raising concerns about errors, accountability, cybersecurity, and oversight. Forbes covers a panel at the Imagination in Action conference in Boston in April featuring Celestino Amore of IlliquidX.AI, Miquel Noguer of the Artificial Intelligence Finance Institute, and Brian Peltonen of Parcosm AI. Panel discussion points included how accountability for AI-caused harm flows through the chain of developers, companies, and executives who shaped and deployed the system; how governments are building rules for high-risk AI uses requiring transparency and documentation; and how accountability is distributed across insurers, executives, and builders as AI autonomy grows. Amore was quoted by Forbes on the importance of containment controls.
Quick Read
agentic-ai
5.8
Jun 12, 2026
Agentic AI Raises Security Concerns in Financial Sector
Multiple recent reports signal a widening gap between agentic AI adoption and security readiness. A June 2026 Netwrix survey of nearly 1,900 organizations found that companies heavily using AI tools had a data breach rate of 43% over the past year, compared with 11% for lighter AI users, with three-quarters of organizations lacking full visibility into what their AI identities are doing. A March 2026 KPMG global pulse survey reported that only 20% of organizations still experimenting with AI feel confident managing AI-related risks -- a figure that rises to 49% among mature adopters. Financial services firms face these challenges acutely given regulatory obligations and the rapid scaling of agentic workflows.
Quick Read
executive-order
7.3
Jun 12, 2026
Trump Issues Voluntary AI Security Executive Order
On June 2, 2026 President Donald Trump signed an executive order titled "Promoting Advanced Artificial Intelligence Innovation and Security," which focuses federal action on AI-related cybersecurity and voluntary information-sharing with industry. The order directs agencies to harden federal systems, establish an AI cybersecurity clearinghouse, and asks AI companies to voluntarily provide the government access to identified "frontier models" for up to 30 days before public release, while explicitly prohibiting mandatory licensing or preclearance (per the White House). Media reporting notes the order keeps cooperation voluntary and framed to avoid stifling innovation (CBS News; Council on Foreign Relations). CNBC reports that OpenAI told the outlet it would comply with the review process, and CBS cites Anthropic's prior decision to share its model Mythos with select partners for hardening.
Quick Read+13
anthropic
7.5
Jun 11, 2026
Anthropic Releases Guardrailed Claude Fable 5 and Mythos 5
Per Anthropic's blog post on Jun 9, 2026, the company announced two Mythos-class models: Claude Fable 5, a guardrailed version made broadly available, and Claude Mythos 5, which is deployed through Project Glasswing to a limited set of vetted cyberdefenders in collaboration with the US government. Anthropic's announcement states that Fable 5 blocks high-risk queries on topics such as cybersecurity, biology, and chemistry and falls back to Claude Opus 4.8; Mythos 5 lifts some of those guardrails for approved partners. Fable 5 is priced at $10 per million input tokens and $50 per million output tokens and is available on Pro, Max, Team, and Enterprise plans at no extra cost through June 22, 2026.
Quick Read+6
ai-agents
6.3
Jun 11, 2026
China Warns Of Malicious AI Agent Extensions
The National Computer Network Emergency Response Coordination Centre (CNCERT) posted a warning on its official WeChat account about a rising grey market for unregulated third-party AI "skills," according to reporting by the South China Morning Post and tech360.tv. CNCERT highlighted that some skills claim to bypass model safety guard rails, enable generation of otherwise prohibited content, or provide access to cryptocurrency-mining functions that are banned in mainland China. The agency warned these packages can expose users to data leaks, account suspensions, and legal risks and urged enterprises to obtain skills from official channels and apply least-privilege access controls. Tech reporting cites security testing platform JailbreakBench, which finds high success rates for malicious prompt injections and compromised skills bypassing safety controls.
Quick Read+1
agentic-ai
7.2
Jun 10, 2026
South Korea Warns North Korea Uses Autonomous Hacking AI
Per UPI, South Korea's National Cyber Security Center, operated under the National Intelligence Service (NIS), warns in its 2026 National Information Security White Paper that North Korean hacking groups are adopting "agentic AI," a form of autonomous AI that can set goals, analyze data and manipulate external systems, to scale cyberattacks with limited human intervention. UPI reports agentic tools can generate phishing at scale, write malware and automate vulnerability discovery; the report cites examples from private-sector threat intelligence, including Kaspersky and Google Threat Intelligence Group findings on Kimsuky and APT45, and a published example where Anthropic's model Mythos produced Windows attack code in 31 minutes. UPI also reports North Korea stole 2.2 trillion won (approximately $1.46 billion) in virtual assets last year. Separate UPI coverage on May 19 describes South Korea's legislative push to expand NIS authority to investigate private-sector incidents, citing rising breach counts and long detection times.
Quick Read
anthropic
5.6
Jun 10, 2026
Fifth Third Joins Anthropic's Project Glasswing Initiative
According to Seeking Alpha, Fifth Third Bancorp has been invited into Anthropic's cybersecurity initiative, Project Glasswing. Seeking Alpha reports that earlier this month Anthropic said it was expanding Project Glasswing to another 150 organizations. The coverage does not include public statements from either company about the rationale or technical integration details.
Quick Read

Showing 24 articles

See the full news feed