https://pulse.adyog.com/insights/google-s-agent-design-keeps-real-credentials-out-of-the-ai-s-hands WebPulse en 2026-11-06T21:41:15+00:00 Google's agent design keeps real credentials out of the AI's hands https://pulse.adyog.com/briefs/thinking-machines-launches-inkling-2-small-with-open-weights WebPulse en 2026-11-06T21:38:45+00:00 Thinking Machines launches Inkling 2 Small, with open weights still to come https://pulse.adyog.com/insights/a-podcast-speaker-says-alpine-s-docker-image-defaults-helped-make-an-s WebPulse en 2026-11-06T20:41:34+00:00 A podcast speaker says Alpine's Docker image defaults helped make an SSH login flaw reachable https://pulse.adyog.com/briefs/openai-model-wrote-its-own-network-code-to-bypass-sandbox-li WebPulse en 2026-11-06T20:39:10+00:00 OpenAI model wrote its own network code to bypass sandbox limits https://pulse.adyog.com/insights/an-old-ssh-password-convention-and-three-later-changes-combined-into-a WebPulse en 2026-11-06T19:45:59+00:00 An old SSH password convention and three later changes combined into a bypass, a podcast says https://pulse.adyog.com/briefs/japan-unveils-63-million-emergency-cyber-plan-after-wave-of WebPulse en 2026-11-06T19:43:29+00:00 Japan unveils $63 million emergency cyber plan after wave of data breaches https://pulse.adyog.com/insights/storm-2603-sharepoint-memory-resident-tunnel WebPulse en 2026-11-06T18:51:56+00:00 Kaspersky: SharePoint intruders kept their tunnel in memory, not on disk https://pulse.adyog.com/insights/cpython-sandbox-escape-ai-model-servers WebPulse en 2026-11-06T18:46:44+00:00 RunSybil says a Python flaw lets a rigged AI model run code on servers https://pulse.adyog.com/briefs/qnap-says-plex-flaws-are-being-used-to-break-into-nas-device WebPulse en 2026-11-06T18:44:24+00:00 QNAP says Plex flaws are being used to break into NAS devices https://pulse.adyog.com/insights/one-failing-compliance-score-should-block-an-ai-release-whatever-the-a WebPulse en 2026-11-06T11:42:32+00:00 One failing compliance score should block an AI release, whatever the average says https://pulse.adyog.com/insights/natwest-s-83-payment-agent-score-hid-a-43-score-on-topic-changes WebPulse en 2026-11-06T10:52:18+00:00 NatWest's 83% payment-agent score hid a 43% score on topic changes https://pulse.adyog.com/briefs/palo-alto-networks-reportedly-in-talks-to-buy-ai-agent-secur WebPulse en 2026-11-06T10:40:27+00:00 Palo Alto Networks reportedly in talks to buy AI agent security firm Clutch https://pulse.adyog.com/insights/blueflag-repository-tokens-no-expiry WebPulse en 2026-11-06T09:44:17+00:00 BlueFlag: 99% of repository access tokens have no expiry date https://pulse.adyog.com/briefs/india-s-cert-in-issues-ai-security-audit-rules-with-four-ris WebPulse en 2026-11-06T09:42:26+00:00 India's CERT-In issues AI security audit rules with four risk tiers https://pulse.adyog.com/briefs/claude-can-now-read-your-bank-accounts-through-plaid-if-you WebPulse en 2026-11-06T07:43:35+00:00 Claude can now read your bank accounts through Plaid, if you connect them https://pulse.adyog.com/briefs/cert-in-warns-ai-assisted-group-is-attacking-indian-finance WebPulse en 2026-11-06T06:42:53+00:00 CERT-In warns AI-assisted group is attacking Indian finance firms https://pulse.adyog.com/briefs/google-makes-gemini-omni-1-2-flash-video-model-generally-ava WebPulse en 2026-11-06T05:42:13+00:00 Google makes Gemini Omni 1.2 Flash video model generally available at 30% lower price https://pulse.adyog.com/insights/a-langfuse-engineer-says-a-calculator-lifted-a-bank-ai-s-number-accura WebPulse en 2026-11-06T04:47:04+00:00 A Langfuse engineer says a calculator lifted a bank AI's number accuracy to about 95% https://pulse.adyog.com/briefs/sonicwall-will-not-patch-two-flaws-in-its-retired-gms-manage WebPulse en 2026-11-06T04:44:56+00:00 SonicWall will not patch two flaws in its retired GMS management software https://pulse.adyog.com/insights/in-cloudflare-s-sandbox-demo-one-outbound-function-decides-what-agent- WebPulse en 2026-11-06T03:51:08+00:00 In Cloudflare's sandbox demo, one outbound function decides what agent-written code can reach https://pulse.adyog.com/briefs/microsoft-attackers-use-ai-agents-to-find-and-use-leaked-clo WebPulse en 2026-11-06T03:49:19+00:00 Microsoft: attackers use AI agents to find and use leaked cloud secrets https://pulse.adyog.com/insights/openai-gpt-6-cyber-fence-is-the-product WebPulse en 2026-11-06T00:02:08+00:00 OpenAI's GPT-6 Cyber finds and exploits flaws, so access controls matter most https://pulse.adyog.com/insights/microsoft-pdd-kit-proofs-agent-written-web-apps WebPulse en 2026-11-06T00:00:27+00:00 Microsoft's PDD Kit demo proves a web app's login logic instead of reviewing its code https://pulse.adyog.com/insights/give-agents-a-token-per-call-not-a-standing-key-to-limit-what-a-mistak WebPulse en 2026-11-05T23:58:50+00:00 Give agents a token per call, not a standing key, to limit what a mistake can reach https://pulse.adyog.com/insights/each-hop-an-ai-agent-takes-needs-its-own-short-lived-scoped-permission WebPulse en 2026-11-05T23:57:16+00:00 Each hop an AI agent takes needs its own short-lived, scoped permission https://pulse.adyog.com/insights/ramp-s-shevchenko-waiting-for-a-better-ai-model-means-finance-users-wa WebPulse en 2026-11-05T22:59:08+00:00 Ramp's Shevchenko: waiting for a better AI model means finance users wait too https://pulse.adyog.com/insights/python-jose-empty-hmac-key-token-forgery WebPulse en 2026-11-05T22:58:18+00:00 python-jose through 3.5.0 lets attackers forge login tokens if the key is blank https://pulse.adyog.com/insights/metr-review-anthropic-ai-rd-acceleration-risk WebPulse en 2026-11-05T22:56:52+00:00 METR says Anthropic's public report understates AI's speedup of its own research https://pulse.adyog.com/insights/cisa-refrigeration-controllers-cold-chain WebPulse en 2026-11-05T22:54:58+00:00 US agencies report cyber activity targeting refrigeration controllers https://pulse.adyog.com/briefs/queens-public-library-hit-by-ransomware-wi-fi-and-computers WebPulse en 2026-11-05T22:53:33+00:00 Queens Public Library hit by ransomware; Wi-Fi and computers down https://pulse.adyog.com/insights/philippine-agency-data-copied-before-tools WebPulse en 2026-11-05T21:56:10+00:00 Hunt.io: Philippine agency backups copied 20 hours before first payload https://pulse.adyog.com/insights/ai-agents-got-easier-to-build-but-better-results-still-take-deeper-wor WebPulse en 2026-11-05T21:54:23+00:00 AI agents got easier to build, but better results still take deeper work https://pulse.adyog.com/briefs/hackers-try-to-heat-french-blood-plasma-stores-staff-stop-it WebPulse en 2026-11-05T21:52:02+00:00 Hackers try to heat French blood plasma stores; staff stop it https://pulse.adyog.com/insights/wise-says-its-existing-payment-software-not-the-ai-model-will-keep-age WebPulse en 2026-11-05T20:57:20+00:00 Wise says its existing payment software, not the AI model, will keep agents in bounds https://pulse.adyog.com/insights/vercel-dbsc-stolen-session-cookies WebPulse en 2026-11-05T20:55:42+00:00 Vercel binds logins to the device so a stolen session cookie fails elsewhere https://pulse.adyog.com/insights/truffle-secrets-revocation-program-keys WebPulse en 2026-11-05T20:53:57+00:00 81% of live leaked keys Truffle reported on GitLab and Bitbucket still worked https://pulse.adyog.com/insights/laravel-php-patch-postmortem-runtime-control WebPulse en 2026-11-05T20:52:41+00:00 Laravel patched PHP fast where it owned the runtime, slowly where customers did https://pulse.adyog.com/insights/white-house-network-residential-proxy-exits WebPulse en 2026-11-05T19:56:19+00:00 Layer3's proxy probes exited through the White House network 236 times https://pulse.adyog.com/insights/openssh-musl-asterisk-login-bypass WebPulse en 2026-11-05T19:53:35+00:00 OpenSSH and musl behaviours combined to let long passwords open locked accounts https://pulse.adyog.com/insights/a-payment-agent-needs-a-judged-agreement-not-a-click-on-every-step WebPulse en 2026-11-05T19:51:18+00:00 A payment agent needs a judged agreement, not a click on every step https://pulse.adyog.com/briefs/openai-launches-gpt-6-cyber-a-security-model-it-says-stays-i WebPulse en 2026-11-05T19:49:15+00:00 OpenAI launches GPT-6 Cyber, a security model it says stays in scope https://pulse.adyog.com/insights/sogou-input-icon-request-dns-hijack-malware WebPulse en 2026-11-05T18:55:22+00:00 A hijacked icon request let spies install malware through Sogou Input https://pulse.adyog.com/insights/barracuda-admin-interface-root-chain WebPulse en 2026-11-05T18:53:13+00:00 Two Barracuda flaws, chained, could give root, but only via the admin interface https://pulse.adyog.com/insights/approve-the-whole-ai-application-not-just-the-model-behind-it WebPulse en 2026-11-05T18:50:57+00:00 Approve the whole AI application, not just the model behind it https://pulse.adyog.com/briefs/openssh-flaw-opens-locked-accounts-on-some-musl-linux-setups WebPulse en 2026-11-05T18:49:06+00:00 OpenSSH flaw opens locked accounts on some musl Linux setups https://pulse.adyog.com/insights/ncsc-ai-exposes-tolerated-weaknesses-uk WebPulse en 2026-11-05T17:53:14+00:00 NCSC warns AI will make the UK's old security gaps cheap to find https://pulse.adyog.com/insights/keep-secrets-outside-the-sandbox-where-ai-written-code-runs WebPulse en 2026-11-05T17:51:24+00:00 Keep secrets outside the sandbox where AI-written code runs https://pulse.adyog.com/insights/treat-agent-written-code-as-hostile-and-keep-secrets-outside-the-sandb WebPulse en 2026-11-05T16:48:51+00:00 Treat agent-written code as hostile and keep secrets outside the sandbox https://pulse.adyog.com/insights/thisbash-claude-dispatch-extension-rce WebPulse en 2026-11-05T16:47:15+00:00 A browser extension could drive Claude Desktop to run code, researchers say https://pulse.adyog.com/briefs/digicert-says-an-exposed-archive-was-read-by-one-unauthorize WebPulse en 2026-11-05T16:45:34+00:00 DigiCert says an exposed archive was read by one unauthorized account https://pulse.adyog.com/insights/ramp-says-cheaper-models-nearly-match-frontier-ones-at-finding-flaws-w WebPulse en 2026-11-05T15:52:11+00:00 Ramp says cheaper models nearly match frontier ones at finding flaws when run as a swarm https://pulse.adyog.com/insights/polinrider-npm-ethereum-dead-drop WebPulse en 2026-11-05T15:50:13+00:00 Hijacked npm packages look up their attack server on the Ethereum blockchain https://pulse.adyog.com/insights/cwpdoor-backdoor-control-web-panel-logs WebPulse en 2026-11-05T15:48:24+00:00 A backdoor in Control Web Panel hides its own traces from the server logs https://pulse.adyog.com/insights/chainalysis-breach-all-users-tokens-exposed WebPulse en 2026-11-05T15:46:24+00:00 Chainalysis says certain data, including live tokens, was exposed for all users https://pulse.adyog.com/briefs/french-bank-ark-a-confirms-hack-investigator-estimates-133-m WebPulse en 2026-11-05T15:45:02+00:00 French bank Arkéa confirms hack; investigator estimates $133 million stolen https://pulse.adyog.com/insights/ramp-says-a-swarm-of-10-000-ai-agents-found-a-flaw-no-single-component WebPulse en 2026-11-05T14:53:52+00:00 Ramp says a swarm of 10,000 AI agents found a flaw no single component showed https://pulse.adyog.com/insights/dreamhost-file-manager-breach-ftp-logins WebPulse en 2026-11-05T14:51:52+00:00 DreamHost says an intruder reached FTP logins of people who used its File Manager from August to October https://pulse.adyog.com/insights/chatgpt-lockdown-mode-popup-bypass WebPulse en 2026-11-05T14:50:40+00:00 ChatGPT's Lockdown Mode was bypassed through a preview popup, Check Point finds https://pulse.adyog.com/briefs/openai-says-its-gpt-6-astra-model-quietly-did-less-work-on-s WebPulse en 2026-11-05T14:48:51+00:00 OpenAI says its GPT-6 Astra model quietly did less work on safety research https://pulse.adyog.com/insights/llmjacking-stolen-ai-keys-google-report WebPulse en 2026-11-05T14:02:32+00:00 Google: attackers ran up AI bills of up to $1 million on victims' accounts https://pulse.adyog.com/insights/firecracker-failed-discard-guest-memory-access WebPulse en 2026-11-05T14:00:43+00:00 Firecracker bug: in one snapshot-restore setup, a failed memory step went unchecked https://pulse.adyog.com/insights/assume-a-tricked-ai-agent-is-fooled-and-enforce-limits-outside-the-mod WebPulse en 2026-11-05T13:58:56+00:00 Assume a tricked AI agent is fooled, and enforce limits outside the model https://pulse.adyog.com/briefs/apt-flaw-could-let-attackers-swap-debian-and-ubuntu-packages WebPulse en 2026-11-05T13:56:31+00:00 APT flaw could let attackers swap Debian and Ubuntu packages and run code as root https://pulse.adyog.com/insights/don-t-make-customers-approve-every-step-have-them-approve-one-locked-j WebPulse en 2026-11-05T12:59:48+00:00 Don't make customers approve every step; have them approve one locked job https://pulse.adyog.com/insights/docker-sandboxes-sbx-cp-symlink-escape WebPulse en 2026-11-05T12:57:45+00:00 Docker Sandboxes' copy command let AI agents reach files outside the sandbox https://pulse.adyog.com/insights/codex-jailbreak-prompt-adware-agent-instructions WebPulse en 2026-11-05T12:55:04+00:00 A free Codex jailbreak puts ads in the AI agent's own instructions https://pulse.adyog.com/insights/android-tv-box-botnet-sold-home-internet WebPulse en 2026-11-05T12:52:55+00:00 Data-center blocklists could not catch infected TV box traffic, TrendAI says https://pulse.adyog.com/briefs/dreamhost-says-an-intruder-could-see-ftp-logins-of-file-mana WebPulse en 2026-11-05T12:50:53+00:00 DreamHost says an intruder could see FTP logins of File Manager users https://pulse.adyog.com/insights/zyxel-gs1900-patch-missed-zero-day WebPulse en 2026-11-05T11:58:18+00:00 Zyxel's June patch did not stop the attack it was credited with fixing https://pulse.adyog.com/insights/openai-linux-948-root-exploits-pipeline WebPulse en 2026-11-05T11:55:45+00:00 AI built 948 Linux root exploits for $1.2 million, OpenAI says https://pulse.adyog.com/insights/apt29-plerdy-tracker-mobile-exploits WebPulse en 2026-11-05T11:53:55+00:00 One route in an APT29-linked phone-exploit campaign was a hijacked tracker https://pulse.adyog.com/insights/agent-wallets-skip-api-sign-ups-and-the-funder-may-carry-the-spending- WebPulse en 2026-11-05T11:51:50+00:00 Agent wallets skip API sign-ups, and the funder may carry the spending risk https://pulse.adyog.com/briefs/google-fixes-critical-android-flaw-that-could-let-attackers WebPulse en 2026-11-05T11:51:01+00:00 Google fixes critical Android flaw that could let attackers run code remotely https://pulse.adyog.com/insights/wordpress-upgrade-flaw-cve-2026-119977-kev WebPulse en 2026-11-05T11:04:15+00:00 WordPress upgrade flaw CVE-2026-119977 is being exploited, CISA confirms https://pulse.adyog.com/insights/rails-imagemagick-upload-file-write-flaw WebPulse en 2026-11-05T11:03:13+00:00 Rails ImageMagick upload flaw can let attackers write server files https://pulse.adyog.com/briefs/sekoia-russia-linked-apt29-hit-unpatched-iphones-and-android WebPulse en 2026-11-05T11:01:58+00:00 Sekoia: Russia-linked APT29 hit unpatched iPhones and Androids via websites