←
AgentDNA341 posts
AgentDNA avatar

AgentDNA

@Agent_DNA

Redefining digital trust. AgentDNA encodes authenticity, governance, and context into every intelligent entity

◷ Joined November 2025▧ 5 Photos and videos
KC avatar
⟳ AgentDNA retweeted
KC@KCRubix·Nov 7

35% breaches began with phishing or stolen credentials. Almost all breaches involve stolen credentials at some stage. OTPs, API keys, SSH keys... Passwords is the root problem. Harden and airgap your auth now with #XID, #AuthAMS and @ThinC_AUTH. @Agent_DNA

a16z@a16zNov 6

Some of the biggest openings in security have nothing to do with patching How breaches began, per a 2026 report: - 31% exploited a vulnerability - 35% used phishing, stolen logins or impersonation scams And nearly half now involve a vendor or partner, up 3x since 2024 Full piece from a16z's Asaf Ezra and Steph Zhang on six cybersecurity companies worth building: a16z.news/p/defense-in-depth…

Post attachment
KC avatar
⟳ AgentDNA retweeted
KC@KCRubix·Nov 8

Also need to build a graph based system to be able to link all the agents that are acting in concert. @Agent_DNA

KC avatar
⟳ AgentDNA retweeted
KC@KCRubix·Nov 5

Real concern. Only way is to be able to real time link all the agents who need add the card to be able to block at run time. @Agent_DNA

Imagine adding a stolen card to 100 AI agents and having them all shop at once before the issuer could block them...

KC avatar
⟳ AgentDNA retweeted
KC@KCRubix·Nov 5

All bank admin accounts need to harden and air gap access immediately. Including for SSH and session keys. @EnsurityTech #XID and @ThinC_AUTH will help mitigate significantly.

‼️🇮🇳𝗖𝗘𝗥𝗧-𝗜𝗻 𝗪𝗮𝗿𝗻𝘀 𝗼𝗳 𝗘𝘀𝗰𝗮𝗹𝗮𝘁𝗶𝗻𝗴 𝗔𝗜-𝗗𝗿𝗶𝘃𝗲𝗻 𝗖𝘆𝗯𝗲𝗿𝗮𝘁𝘁𝗮𝗰𝗸𝘀 𝗯𝘆 𝗖𝗥𝗔𝗙𝗧𝗬 𝗙𝗔𝗟𝗖𝗢𝗡 𝗧𝗮𝗿𝗴𝗲𝘁𝗶𝗻𝗴 𝗜𝗻𝗱𝗶𝗮'𝘀 𝗙𝗶𝗻𝗮𝗻𝗰𝗶𝗮𝗹 𝗦𝗲𝗰𝘁𝗼𝗿 India’s CERT-In issued a critical advisory warning of an ongoing AI-assisted cyberattack campaign attributed to CRAFTY FALCON, targeting financial institutions, fintech firms, payment providers, and cryptocurrency exchanges. Using AI-powered tools such as ARTEX, PentAGI, Cairn, and CyberStrikeAI, the group exploits system vulnerabilities, compromises payment APIs, and steals funds and cryptocurrency. Attackers reportedly conduct INR 1 test transactions before larger unauthorized transfers, often striking during weekends and holidays. Some organizations have been compromised again following incomplete remediation. For more insights into the latest cyberattacks worldwide, visit HackRisk.io

Post attachment
KC avatar
⟳ AgentDNA retweeted
KC@KCRubix·Nov 5

Several banks around the world reporting failures in access to inter bank transfer systems leading to customer account drainage. Banks must harden and airgap their core banking and inter bank transfer access systems asap. @EnsurityTech @ThinC_AUTH #XID #AuthAMS

【ついにサイバー攻撃が銀行本体にも及んだ!】 アプリは開いていない 認証の通知も来ていない なのに、口座から10万台湾ドル(約50万円)が消えた 11月3日、台湾の大手「永豐銀行」で418人・427件、総額4176万台湾ドル(約2億円)が一斉に抜き取られました ▼11月3日(台湾時間) 9時台 不正送金が始まる 9:29 「勝手に引き出された」と苦情が入り始める 10:30 送金先の富邦銀行口座の凍結を依頼 12:00 銀行が「重大事件」と判断 12:50 金融当局に通報 13:47 原因を特定し新アプリを停止、攻撃が止まる 同日中 被害額を全額返金 ▼11月4日 銀行が「外部ハッカーによる攻撃」と正式に認め、会長と社長が謝罪 台北地検が捜査に着手 ■お名前.comの件とは"質"がちがう 先月のお名前.com デスクトップクラウドの事件 あれはFX会社や証券会社そのものではなく、取引ツールを動かす"外側"のサーバーが破られ、そこに保存されていたログイン情報で口座が操られたものでした いわば外部連携の入口が狙われた話で、金融機関そのものが侵入されたわけではありません 今回はちがいます ✅狙われたのは銀行自身のネットバンキング取引システム 被害者はパスワードを盗まれたわけでも、怪しいリンクを踏んだわけでもない 銀行の"玄関"そのものがこじ開けられ、本人のログインも認証もないまま送金が実行されたのです ■手口 銀行によると、新システムを作った外部ベンダーのプログラムの欠陥と設定を突かれ、ハッカーはIDを総当たり的に試して有効な口座を探し当てた 被害者が全員、身分証番号が「P」(雲林県)で始まる人だったのはそのためで、個人情報の流出は今のところ確認されていないとしています ■いちばん不気味な点 送金先は、本人も何も知らなかった70代の元教師の口座 お金は引き出されていません 最初は少額から試し、やがて1口座10万元に"精度"を上げていった形跡もあり、警察は「技術を見せつけるのが目的では」とみています ✅裏を返せば、その気になれば本当に奪えたということです ■他人事ではない 銀行アプリや取引システムを外部ベンダーが作るのは日本でも珍しくありません 「銀行だから安全」はもう通用しない 私見では(GMOもですが)PRSサイオプのショーと考えます おそらく規制強化を目的とするものでしょうね

KC avatar
⟳ AgentDNA retweeted
KC@KCRubix·Nov 5

Customers controlling access to their accounts with hardend and air gapped access with @ThinC_AUTH and #XID should be allowed to let their agents access banking systems. @Agent_DNA

.@WellsFargo appears to be the first bank to start requiring customers to prove they've turned off their #AIagents' access to their #bankaccounts, or else be liable for any rogue AI agent activity. More here: sbrew.link/DSskXedd

KC avatar
⟳ AgentDNA retweeted
KC@KCRubix·Oct 29

Identity is the biggest moat in cyber security. Human as well as agentic. #KYH #KYC

KC@KCRubixOct 28

Cloud desktops used by #GMOInternet by many Japanese FX trading houses were breached and users passwords compromised. Switch from #passwords to secure biometric fido access urgently. @EnsurityTech AuthAMS with @ThinC_AUTH will help orgs to avoid losses. #Japan

KC avatar
⟳ AgentDNA retweeted
KC@KCRubix·Oct 25

More than a quarter of people doing shopping with #Muse - that seems big for a month old app. If this share sustains with larger penetration to be seen, but agentic commerce appears a real deal. Need to fix agentic security though at first!

Shopping is the most common use case for Muse And the average transaction size is $200

KC avatar
⟳ AgentDNA retweeted
KC@KCRubix·Oct 25

Agree, banks have a big fraud and liability issue to address. Some big unanswered questions: 1. If the user denies the agent action, how does bank prove the user did give consent. 2. How does the bank prove user gave the credential willingly and not misused by the agent. 3. Is the agent a legitimate rep of the user or a hacker who stole credentials 4. Is the intended action approved by the user even if credential shared. How is the intent verified dynamically? 5. Agents by design would like to have the broadest intent (even users may be ok), but how does bank enforce explicit auth from user when it wants. 6. Did the agent create sub agents to act ? 7. Is the agent prompted to do a malicious activity. Even as banks are trying to deal with agentic actions, AI is making passwords and biometrics themselves highly risky. We need to ensure the hardened user root of trust (eliminate passwords) to pass onto agents and enforce continuous auth. And treat agentic actions as not simple key or x406 certain based auth, but derived from continuous DNA. KYA and KYH will need to replace KYC.

Your agent looks like a giant fraud and liability problem to your bank. Even if you sent it. The bank needs to know who it’s acting for and whether you approved what it’s doing. I mapped the protocols that could make this work:

KC avatar
⟳ AgentDNA retweeted
KC@KCRubix·Oct 24

Anecdotal reports of similar losses at multiple banks in the recent weeks - does not look isolated to Central Bank. This theft seems to be breach of core banking access controls at the bank end. Banks must immediately assess their core banking access controls and shift away from password/OTP/api key based legacy systems. @EnsurityTech

India news: Central Bank of India lost ₹13.96 crore in 187 unauthorised RTGS transfers to 36 banks from Oct 15 to 17, police say. Unknown people allegedly got into the RBI-owned IFTAS Central SFMS server through the bank's core banking system. The money was customers' funds. The bank first called it a "technical glitch". RBI, CERT-In and NCIIPC have been told and Maharashtra cyber police have a case. If you bank there, check your RTGS debits from those dates. hindustantimes.com/cities/mu…

KC avatar
⟳ AgentDNA retweeted
KC@KCRubix·Oct 23

Agents appearing to act beyond users instructions is not itself a breach. It is just that tools dont exist to bind agents to users for limited scope. We have been building control mechanisms with @Agent_DNA. However, since agents like #Muse have access to API keys and passwords of users, hackers can steal them to attack banks. This is a real and big risk.

The situation is getting worse every hour. AI agents are entering U.S. bank accounts, and banks are scrambling to shut them out. Axios reports that some appear to attempt actions beyond users instructions, raising fears over who is really in control.

Post attachment
KC avatar
⟳ AgentDNA retweeted
KC@KCRubix·Oct 22

Agents are currently using users credentials and API keys to access. This must stop. Agents cant use users credentials, they must have own ID and get scoped auth permissions from users. @Agent_DNA Chain of Custody Auth ensures agents cant have sweeping permissions.

JUST IN: 🇺🇸 Morgan Stanley, Bank of America, JPMorgan, and Wells Fargo detect unauthorized AI agents accessing customer accounts and attempting to act without user instructions.

Post attachmentPost attachmentPost attachmentPost attachment