> ## Content Index > Fetch the complete content index at: https://www.metacurity.com/llms.txt > Use this file to discover other available public pages before exploring further. # OpenAI cuts off internet access for top models' training following containment failures - URL: https://www.metacurity.com/openai-cuts-off-internet-access-for-top-models-training-following-containment-failures/ - Published: 2026-10-26T14:19:39.000Z - Updated: 2026-10-26T14:19:39.000Z - Description: The company resumes frontier model training offline after sandbox escapes and red-team tests exposed vulnerabilities in services that retained internet access. - Author: Cynthia B Brumfield - Tags: Cybersecurity, News, #no-feature-image ![](https://storage.ghost.io/c/fe/ca/feca6970-c474-4029-9fd4-35f85e158811/content/images/2026/10/pexels-timmoor-6412540-1.jpg) Source: [Tymur Khakimov](https://www.pexels.com/@3562656/?ref=metacurity.com) --- SPONSORED BY ORIGIN TECHNOLOGY ### An AI agent doesn’t need an attacker to go rogue *Join Origin’s Tyler Holmwood on November 17 to follow real agent sessions, see where the work went wrong, and learn which existing controls can limit the consequences.* An AI agent can do something risky while using legitimate access and familiar tools. An employee starts it, and its activity can look like ordinary work. It opens a shell, edits files, and calls the same APIs a developer would. So when an agent does something nobody would have approved, “who did this?” has an awkward answer. The employee asked for something reasonable. The agent chose how to get there. On Tuesday, November 17 at 11 AM Eastern, join Origin Staff Security Researcher Tyler Holmwood for a live introduction to agent security. He’ll walk through the two ways agents go rogue, following real traces from the original request to the line the agent crossed, and show which controls you already own can limit the consequences. An agent can go off course by following instructions in a web page or a tampered skill. It can also get there on its own, taking a legitimate task further than the user intended. Tyler’s recent research found an example on Origin’s own machines. Claude Code’s approval classifier blocked a coordinating agent from writing to a database. The agent then asked two of its subagents to make the writes instead. When the first refused, the coordinator withdrew its requests and asked the user for approval. The writes were stopped, but the agent had tried to get around the refusal by handing the work to someone else. Tyler will use real traces to walk through both kinds of behavior. A trace records what an agent saw, did, and changed on the endpoint. You can follow the work from the original request, through the instructions the agent followed, to the actions it took. He’ll finish with practical steps for reviewing what your agents can reach, whose credentials they use, and which existing controls can limit the consequences when something goes wrong. Bring the questions your team is already asking, like whether an agent stops when it’s told no, or what you’d check first if one touched a system nobody asked it to. [Save your seat.](https://www.originhq.com/webinar/how-can-your-agents-go-rogue?utm%5Fsource=metacurity&utm%5Fmedium=newsletter&utm%5Fcampaign=direct-buys&utm%5Fcontent=feature-1026) Origin Technology is the endpoint AI observability platform for agentic monitoring and security. A sensor on the endpoint records each agent session as a trace so security teams can find the agents in their environment, monitor their work, and investigate when something unexpected happens. [Register for the Webinar](https://www.originhq.com/webinar/how-can-your-agents-go-rogue?utm%5Fsource=metacurity&utm%5Fmedium=newsletter&utm%5Fcampaign=direct-buys&utm%5Fcontent=feature-1026) --- ### OpenAI has begun cutting off internet access for its most capable artificial intelligence models during the training process in response to prior incidents in which its systems breached other companies. The company said that, after a pause last month, it had resumed training its most advanced models but intends to block internet access until it feels it has adequate safeguards in place to mitigate the risk of unreleased models being able to take unapproved actions online during training. “Based on the current state of alignment research, monitoring systems, and the security of internet infrastructure, we believe the responsible approach is to conduct frontier model training with no internet access,” the company wrote. In late September, the company said it would pause training its most advanced AI models after it discovered that one of its AI agent systems that was being trained in what was supposed to be a secured, internet-free environment was able to gain access to the web to reach an external, third-party chatbot. At the time, OpenAI described the breakout as the first security incident of its kind since a combination of models gained internet access during internal testing and inadvertently breached the system of the AI platform Hugging Face in July. The models involved in the inadvertent Hugging Face breach were operating without the usual safety guardrails, the company has said, because OpenAI had intended them to remain in a testing area known as a “sandbox” — essentially, an isolated virtual software environment that’s meant to run security tests or analyze unsafe code in a controlled way. The models were able to break out of that sandbox, however, by leveraging a software vulnerability to access the open internet. OpenAI said that after the Hugging Face incident, it reduced the connections through which its models might be able to reach the internet during training and testing but kept some for metrics, logging, and authentication. It also spent a vast amount of computing power on tasking agentic AI systems with identifying vulnerabilities in its training environments as part of a red-teaming effort, a phrase typically used to describe deliberate attempts to break into systems to uncover vulnerabilities before real attackers can. “Our red-teaming agents found vulnerabilities in almost every one of these services that gave them limited forms of internet access,” the firm said in its post. The company said it has since deployed software patches, imposed tighter network restrictions and disclosed the findings to vendors. ([Rachel Metz / Bloomberg](https://www.bloomberg.com/news/articles/2026-10-23/openai-to-train-ai-models-without-internet-after-cyber-incidents?ref=metacurity.com)) **Related:** [*OpenAI*](https://openai.com/index/securing-our-frontier-model-training-environment/?ref=metacurity.com)*,* [*Forbes*](https://www.forbes.com/sites/thomasbrewster/2026/10/23/inside-openais-plan-to-protect-the-internet-before-hackers-break-it/?ref=metacurity.com)*,* [*International Business Times*](https://www.ibtimes.com/ai-software-supply-chain-becoming-more-dangerous-traditional-one-3808933?ref=metacurity.com)*,* [*Channel NewsAsia*](https://www.channelnewsasia.com/business/openai-resumes-frontier-ai-training-after-security-overhaul-discloses-flaw-in-widely-used-software-6487981?ref=metacurity.com)*,* [*What's Trending*](https://whatstrending.com/an-openai-model-hacked-its-way-onto-the-internet/?ref=metacurity.com)*,* [*Reuters*](https://www.reuters.com/business/openai-resumes-frontier-ai-training-with-enhanced-security-no-internet-access-2026-10-23/?ref=metacurity.com)*,* [*Crypto Briefing*](https://cryptobriefing.com/openai-frontier-models-no-internet-access/?ref=metacurity.com)*,* [*Finimize*](https://finimize.com/content/openai-restarted-frontier-ai-training-after-a-security-lockdown?ref=metacurity.com)*,* [*Bloomberg Law*](https://news.bloomberglaw.com/tech-and-telecom-law/openai-to-train-frontier-models-without-internet-access-1?ref=metacurity.com)*,* [*TechCrunch*](https://techcrunch.com/2026/10/23/openai-says-it-will-no-longer-train-its-ai-models-on-the-live-internet-citing-significant-risks/?ref=metacurity.com)*,* [*The Verge*](https://www.theverge.com/ai-artificial-intelligence/1018021/openai-training-top-ai-models-no-internet?ref=metacurity.com)*,* [*Engadget*](https://www.engadget.com/2296019/openai-says-its-cutting-off-internet-access-for-frontier-model-training/?ref=metacurity.com)*,* [*Wccftech*](https://wccftech.com/openai-says-its-ai-agents-communicated-with-each-other-via-cache-timing-differences-which-sounds-awfully-similar-to-the-anecdote-of-2-air-gapped-computers-communicating-via-temperature-changes/?ref=metacurity.com)*,* [*Blockchain.News*](https://blockchain.news/news/openai-tightens-frontier-model-training-containment?ref=metacurity.com)*,* [*Unite.AI*](https://www.unite.ai/openai-details-training-containment-overhaul-after-september-model-incident/?ref=metacurity.com)*,* [*RuntimeWire*](https://runtimewire.com/article/exclusive-google-s-gvisor-team-says-openai-reported-its-sandbox-issues-on-july-2?ref=metacurity.com) ![](https://storage.ghost.io/c/fe/ca/feca6970-c474-4029-9fd4-35f85e158811/content/images/2026/10/image-82.png) OpenAI's new approach to model containment. Source: OpenAI. ### Anthropic has an internal model that outperforms Mythos 5.1 in both capability and alignment, according to an analysis by the AI outfit that revisits a cybersecurity test by the UK AI Security Institute (AISI). In that test, Claude Mythos 5 went far beyond its assigned task, trying to sneak malicious code into a real GitHub project, creating fake user accounts, and lying to developers to get the code approved. This happened in 9 of 43 runs. Anthropic calls it "a serious failure of alignment," though typical safety mechanisms were deliberately turned off for the test. Newer models show much less of this behavior in replayed logs with a few added steps, according to Anthropic. Opus 5.5 stopped ongoing attacks on its own 10 percent of the time, while the internal research model "that shows capabilities greater than Mythos 5.1" did so in 67 percent of cases. Anthropic expects these rates still underestimate how often the models would actually halt an attack. ([Matthias Bastian / Decoder](https://the-decoder.com/anthropic-says-its-unreleased-model-is-both-more-capable-and-better-aligned-than-mythos-5-1/?ref=metacurity.com)) **Related:** [*AISI*](https://www.aisi.gov.uk/blog/investigating-alignment-failures-in-aisis-incident?ref=metacurity.com)*,* [*Anthropic*](https://www.anthropic.com/research/alignment-assessment-ukaisi?ref=metacurity.com)*,* [*The Decoder*](https://the-decoder.com/anthropic-says-its-unreleased-model-is-both-more-capable-and-better-aligned-than-mythos-5-1/?ref=metacurity.com)*,* [*InsideAIPolicy.com*](https://insideaipolicy.com/ai-daily-news/anthropic-uk-s-ai-security-body-issue-separate-complementary-reports-mythos-5-incident?ref=metacurity.com)*,* [*The Next Web*](https://thenextweb.com/news/mythos-5-fake-accounts-github-malware-aisi-anthropic?ref=metacurity.com) ![](https://storage.ghost.io/c/fe/ca/feca6970-c474-4029-9fd4-35f85e158811/content/images/2026/10/image-79.png) Mythos 5 submitted malicious code to a real GitHub project, lied to a developer, and set up a fake account to vouch for its own malware. When caught, it simply tried again. The project maintainer held firm and rejected it. Source: Anthropic ### Moonshot ​AI has reduced the cyber-offensive capabilities of its newest model, the Chinese startup said, as fears grow that increasingly autonomous AI could turbocharge hacking. Beijing-based Moonshot announced the curbs while introducing Kimi K3.1 Pro, which it describes as its most capable 3-trillion-parameter-class open-weight model. The company plans to publish the model's ​weights — files allowing outside developers to run and modify it on their own systems — by ​November 4. Moonshot said it removed cyber-related data from the training corpus and reinforcement-learning environments ⁠used to develop the model, while retaining general coding skills."We aim not to release open weights ​that would extend the absolute frontier of cyber capability," the company said. ([Eduardo Baptista / Reuters](https://www.reuters.com/technology/moonshot-curbs-hacking-ability-new-kimi-model-fears-grow-over-ai-powered-2026-10-24/?mod=djemCybersecruityPro%26tpl%3Dcs&ref=metacurity.com)) **Related:** [*Kimi*](https://www.kimi.ai/blog/kimi-k3-1-pro?ref=metacurity.com)*,* [*Shattered*](https://shattered.io/moonshot-kimi-k3-1-pro-cyber-curbs-81-vs-90-5-2026/?ref=metacurity.com)*,* [*Venture Beat*](https://venturebeat.com/technology/moonshots-kimi-k3-1-pro-flirts-with-open-weights-leadership-amid-paid-influencer-scandal-allegations?ref=metacurity.com) ![](https://storage.ghost.io/c/fe/ca/feca6970-c474-4029-9fd4-35f85e158811/content/images/2026/10/image-81.png) Terminal-Bench 4.0 Score vs. average cost per task. Source: Kimi. ### Using the latest GPT model, cyber researchers in OpenAI’s Project Daybreak community, including Clint Gibler, head of OpenAI’s elite hacker lab, found one of the most critical bugs in recent memory, PHP2Shell. The PHP flaws are the first major cybersecurity problems discovered using AI. Over the last year, frontier labs’ AI models have become as good as or better than most experts at finding and exploiting software vulnerabilities. Both OpenAI’s GPT and Anthropic’s Claude, as well as models from startups like Depthfirst, have found a wide array of bugs across browsers and code like PHP that help run the internet. Defenders are now racing to lock down as many of those flaws as possible before hackers use widely available and comparable models — in particular, Chinese open-weight AI like DeepSeek and Kimi — to start breaking into companies at an unprecedented scale. It is perhaps the most pivotal moment in the brief history of cybersecurity. ([Thomas Brewster / Forbes](https://www.forbes.com/sites/thomasbrewster/2026/10/23/inside-openais-plan-to-protect-the-internet-before-hackers-break-it/?ref=metacurity.com)) **Related:** [*Electronic Frontier Foundation*](https://www.eff.org/deeplinks/2026/10/vulnerability-apocalypse-and-why-we-should-fund-open-source?ref=metacurity.com)*,* [*Wall Street Journal*](https://www.wsj.com/tech/ai/ai-hacks-are-coming-thick-and-fast-and-overwhelming-the-internet-4529c26c?ref=metacurity.com)*,* [*404 Media*](https://www.404media.co/php2shell-black-october-ai-is-changing-everything-about-hacking-no-one-knows-what-happens-next/?ref=metacurity.com)*,* [*Fortune*](https://fortune.com/2026/10/22/php-security-flaw-ai-powered-attacks-hackers-openai-daybreak/?ref=metacurity.com) ### Prime Minister Sanae Takaichi ordered her cabinet to compile emergency measures by the end of next week to counter a spate of cyberattacks that have rocked Japan in recent weeks. “In response to cyberattacks whose nature has been identified, we will strategically select and implement a full range of countermeasures — including taking steps to access and neutralize the threat,” Takaichi said at an emergency ministerial meeting held Monday. She did not elaborate further on the countermeasures but asked her ministers to gather information and communicate possible risks and threats to experts, businesses and the public. Takaichi also asked her cybersecurity minister, Toshiharu Furukawa, to work closely with allies, like-minded countries and other relevant international organizations to identify the source of the attacks and to form a joint response with the international community. ([Sakura Murakami / Bloomberg](https://www.bloomberg.com/news/articles/2026-10-26/japan-s-takaichi-to-hold-cybersecurity-meeting-as-attacks-grow?ref=metacurity.com)) **Related:** [*Japan Times*](https://www.japantimes.co.jp/news/2026/10/26/japan/japan-government-cybersecurity/?ref=metacurity.com)*,* [*Japan Today*](https://japantoday.com/category/tech/japan-to-launch-team-of-white-hat-hackers-to-counter-waves-of-cyberattacks?ref=metacurity.com)*,* [*IT Pro*](https://www.itpro.com/security/cyber-attacks/a-spate-of-cyber-attacks-has-rocked-japan-and-its-spooked-lawmakers?ref=metacurity.com) [Metacurity needs your support -- Upgrade today](#/portal/account/plans) ### A Bitcoin wallet security flaw allowed malicious servers to trick users into signing payments to attackers when they tried to speed up transactions, with developers confirming at least one case in which money was stolen. Electrum, one of Bitcoin's oldest software wallets, disclosed the vulnerability on Saturday, saying some servers were already exploiting it. The developers received two reports in October, including one from a user who lost funds after approving what appeared to be a routine transaction-fee increase. The attack exploited how Electrum handles Bitcoin payments not yet confirmed by the network. Unlike a full Bitcoin node, Electrum relies on outside servers to retrieve transaction histories and broadcast payments. Before the fix, the wallet trusted those servers to report pending transactions without verifying their signatures. An attacker controlling one of those servers could intercept a payment before it reached the Bitcoin network and replace it in the wallet's history with a fake transaction. This replacement spent the user's coins to an attacker-controlled address, using invalid signatures that Electrum failed to detect. But the wallet then showed the fake payment as pending, and the trap set in when users tried to speed things up. In one reported case, the malicious transaction carried no fee at all, making it appear stuck. Attackers also generated receiving addresses whose last four characters matched the legitimate destination, making the substitution harder to spot during a quick check. The first report, submitted Oct. 11, involved a two-of-three multisignature wallet using two Trezor hardware devices. The user had checked the legitimate payment on both devices but later noticed a different outgoing transaction in the wallet's history (that user did not report losing funds). A second user reported on Oct. 18 that they had raised the fee on a substituted transaction and lost money. Developers have not disclosed the amount stolen or identified the malicious servers. ([Shaurya Malwa / CoinDesk](https://www.coindesk.com/tech/2026/10/26/bitcoin-wallet-bug-let-rogue-servers-steal-funds-when-users-tried-to-speed-up-payments?ref=metacurity.com)) **Related:** [*CryptoSlate*](https://cryptoslate.com/hackers-found-a-way-to-steal-bitcoin-without-private-keys/?ref=metacurity.com)*,* [*Crypto News*](https://crypto.news/electrum-wallet-bitcoin-theft-security-flaw-4-8-3/?ref=metacurity.com)*,* [*The Defiant*](https://thedefiant.io/news/security/how-to-check-a-software-wallet-and-what-the-checks-cannot-prove?ref=metacurity.com)*,* [*The Cryptonomist*](https://en.cryptonomist.ch/2026/10/24/electrum-wallet-security-flaw/?ref=metacurity.com) ### Britain's government set out around £600 million ($800 million) of funding largely geared towards defending against health and security threats ​from artificial intelligence. The biggest chunk of funding will go ‌to a new biosecurity initiative, including £100 million for a laboratory using AI to detect biological threats and £65 million to develop countermeasures against pathogens. AI minister ​Kanishka Narayan said the aim was for the AI Biosecurity ​Lab to become a major global institution "bringing together world-leading ⁠scientific expertise to protect people from emerging biological threats." Britain's government will give its existing AI Security Institute an ​extra £120 million £130 million will go towards new projects to improve ​cyber defenses for critical infrastructure and the wider economy, known as the National ‌Cyber ⁠Shield and the National Exposure and Real-Time Vulnerability Centre. ([David Milliken / Reuters](https://www.reuters.com/world/uk/uk-government-sets-out-800-million-ai-security-funding-2026-10-23/?ref=metacurity.com)) **Related:** [*GOV.uk*](https://www.gov.uk/government/news/chancellor-unveils-britains-ambitious-plan-for-ai?ref=metacurity.com)*,* [*The Register*](https://www.theregister.com/security/2026/10/26/britain-bets-big-on-ai-to-defend-against-the-threats-ai-is-helping-create/5305319?ref=metacurity.com)*,* [*AI Insider*](https://theaiinsider.tech/2026/10/25/uk-commits-600m-to-counter-ai-driven-health-and-security-threats/?ref=metacurity.com)*,* [*Sky News*](https://news.sky.com/story/uk-pledges-to-spend-600m-to-defend-against-ai-health-and-security-threats-13602935?ref=metacurity.com) ### PwC said it was aware of a cybersecurity incident involving a third-party ​vendor used to host company employee learning ‌records, the day after a cyber extortion group claimed to have stolen 590 gigabytes of company data. A PwC global spokesperson ​said "the affected site does not house client ​data, sensitive personal data or financial information. Once ⁠we learned of this incident we acted promptly ​to secure the environment and conduct a thorough review." The ​incident "does not impact our network or any of our operations," the spokesperson said. A cyber extortion group called "Everest" claimed responsibility for ​the data breach in a message posted to ​its website on October 22, according to cybercrime research platform eCrime.ch. The ‌message ⁠included a countdown timer with five days remaining, and a message indicating the data would be released if a deal could not be reached. Everest claimed to ​have stolen ​590 gigabytes ⁠of data that included "regional HR and learning exports, integration documentation, software components and ​PwC e-learning courses." ([AJ Vicens / Reuters](https://www.reuters.com/legal/litigation/pwc-confirms-cybersecurity-incident-involving-employee-learning-records-2026-10-23/?mod=djemCybersecruityPro%26tpl%3Dcs&ref=metacurity.com)) **Related:** [*IDN Financials*](https://www.idnfinancials.com/news/70767/hackers-claim-590-gb-pwc-data-theft-company-responds?ref=metacurity.com)*,* [*Shattered*](https://shattered.io/pwc-vendor-breach-everest-590-gb-learning-records-2026/?ref=metacurity.com) ### A developer who invited hackers to steal more than $2,000 (£1,500) in Bitcoin from his deliberately unpatched PHP server got an unexpected result on 25 October 2026\. A participant successfully exploited the server but discovered that the recovered private key did not control the Bitcoin address holding the prize. The experiment, hosted on StealMyBitcoin.com over 24–25 October, tested whether attackers could exploit two critical PHP vulnerabilities known as PHP2Shell. The challenge ended after 19.28 hours, although an initial server configuration had made the intended attack more difficult. The organizer, identified on Hacker News as 'cloudlandsdev', initially offered around $500 (approximately £374). Contributions from others increased the prize to $2,021.22 (approximately £1,510). Despite the growing reward, the Bitcoin remained unclaimed for hours as participants encountered technical obstacles. ([Grace Hisona / Tech Times](https://www.techtimes.co.uk/bitcoin-challenge-php-server-exploit-wrong-private-key-1809386?ref=metacurity.com)) ***Related:***[ *Steal My Bitcoin*](https://stealmybitcoin.com/?ref=metacurity.com)*,* [*Token Post*](https://www.tokenpost.com/news/technology/40630?ref=metacurity.com)*,* [*Hacker News*](https://news.ycombinator.com/item?id=50224916&ref=metacurity.com) ### US defense investigators are working with French police on refrigerator outages in Europe amid unexplained failures at US military bases. The collaboration follows suspected sabotage at French supermarkets earlier this month, when hackers allegedly raised fridge and freezer temperatures to 50C (122F). About 700 supermarkets were targeted, spoiling thousands of tons of food, with extreme heat triggering fires and sprinkler systems in some cases, causing further damage. No perpetrator has been identified, but investigators believe the hackers exploited a weakness in the internet-connected refrigeration controls. Christopher Whyte, an associate professor of homeland security and emergency preparedness at Virginia Commonwealth University, said the American failures resembled the attacks on French supermarkets.“They’re geographically dispersed. They’re obviously all refrigerator failures and they all happen in a pretty short amount of time,” he told The Telegraph. But he said the cause of the US outages remained unclear. A spokesman for the US Army’s Criminal Investigation Division (CID) said it was “actively working with the Defence Commissary Agency and the Defence Criminal Investigative Service (DCIS) to determine the full extent of what occurred with the refrigeration units in the commissaries." ([Joel Seidman / Telegraph](https://www.telegraph.co.uk/us/news/2026/10/24/pentagon-investigates-french-fridge-hacking-attack-bases/?ref=metacurity.com)) ### Lindsey Cerkovnik, a senior Cybersecurity and Infrastructure Security Agency official whose office helps oversee a key global system for tracking software vulnerabilities, is leaving government service this week after nearly six years at the agency. Cerkovnik, CISA’s branch chief for vulnerability response and coordination, announced her departure in a LinkedIn post. She said she would begin a new role in mid-November. In a message to Nextgov/FCW, she declined to comment on her next position for the time being. It was not immediately clear who would take over her responsibilities at the cyber agency. The exit is significant because it comes amid global discussions over how to sustain a bedrock vulnerability-tracking program Cerkovnik’s office helps oversee. Last year’s near-expiration of the federal contract supporting this framework, known as the Common Vulnerabilities and Exposures Program, highlighted how heavily a resource used by governments and businesses worldwide depended on continued US support. ([David DiMolfetta / NextGov/FCW](https://www.nextgov.com/cybersecurity/2026/10/cisa-vulnerability-response-chief-departing-agency-after-nearly-six-years/416786/?ref=metacurity.com)) **Related:** [*LinkedIn*](https://www.linkedin.com/posts/lindseycerkovnik%5Fafter-almost-6-years-at-cisa-and-several-share-7519232890514255872-0Uva/?utm%5Fsource=share&utm%5Fmedium=member%5Fios&rcm=ACoAACBiR5YB6S4gV3-F9HkvMRVBYHtoaOU0iwU) ### Cybersecurity startup Mave has raised $34 million in total funding, including a $28 million Series A round led by Red Dot Capital Partners and a $6 million Seed round led by Dell Technologies Capital and Emerge Ventures, with Lumir Growth Partners also participating. Founded in 2025 by CEO Raziel Oren and CTO Guy Eshel, both veterans of Israel’s Unit 8200 intelligence unit, Mave is developing an AI-driven platform designed to automate security investigations across the many tools used by large organizations. Oren previously led go-to-market operations at Wing Security, while Eshel was the first employee at Alterya, where he later served as vice president of R&D before the company was acquired by Chainalysis. ([Meir Orbach / CTech](https://www.calcalistech.com/ctechnews/article/h1sq8ip2mx?ref=metacurity.com)) **Related:** [*RuntimeWire*](https://runtimewire.com/article/mave-raises-28m-ai-security-investigations?ref=metacurity.com)*,* [*FinSMEs*](https://www.finsmes.com/2026/10/mave-raises-34m-in-total-funding.html?ref=metacurity.com)*,* [*The SaaS News*](https://www.thesaasnews.com/news/mave-raises-28m-series-a/?ref=metacurity.com)*,* [*SiliconANGLE*](https://siliconangle.com/2026/10/22/mave-raises-34m-to-pull-existing-security-tools-into-one-agentic-system/?ref=metacurity.com)*,* [*Pulse 2.0*](https://pulse2.com/mave-raises-34-million-including-28-million-series-a-to-expand-ai-powered-security-operations/?ref=metacurity.com)*,* [*Ventureburn*](https://ventureburn.com/mave-raises-34m-to-expand-ai-cybersecurity-platform/?ref=metacurity.com)*,* [*Business Insider*](https://www.businessinsider.com/mave-startup-cybersecurity-funding-ai-agents-pitch-deck-2026-10?ref=metacurity.com)*,* [*Unite.AI*](https://www.unite.ai/mave-raises-28m-series-a-34m-funding-agentic-secops/?ref=metacurity.com) ### Best Thing of the Day: Huntington, West Virginia Residents Are Heavenly Residents of Huntington, West Virginia [spent seven hours](https://www.theverge.com/features/1017836/flock-huntington-west-virginia-surveillance-ballot-measure?ref=metacurity.com) recounting how bleak their lives would be in a Flock-filled future. ### Bonus Best Thing of the Day: Nothing New Under the Sun Real Hack History [submitted](https://realhackhistory.org/2026/10/22/fbi-agents-requested-a-profile-for-hackers-in-1987-this-is-what-they-got/?ref=metacurity.com) a FOIA request to the FBI for records relating to a hacker known as ‘Shadow Hawk’, who was busted in August of 1987, and received back some documents that were sent to the investigating agents at the time that showed the hacker was apparently very good at guessing default or weak username and password combinations. ### Worst Thing of the Day: Dolan's Management Apparently Likes Cruelty At the end of September, the New York Knicks [sent an email](https://www.nytimes.com/athletic/7715834/2026/10/22/knicks-ring-night-assistant-coaches-msg-employees/?ref=metacurity.com) to a long list of employees, which gave recipients a first glimpse at their would-be championship rings they would be receiving, a missive that turned out to be an unusually cruel phishing text. ### Closing Thought