Oliver Quie (@oli_innerworks) profile picture

Co-founder @InnerworksID & @_redteam_ | Building the Future of Authentication | Destroying All Bots on Twitter Soon | Views my own

Joined April 2024
Photo by Oliver Quie (@oli_innerworks)Photo by Oliver Quie (@oli_innerworks)Photo by Oliver Quie (@oli_innerworks)Photo by Oliver Quie (@oli_innerworks)
Oliver Quie retweeted
"We're not afraid of hyper-intelligent bot swarms on the internet. We use them to optimise our technology." When we launched RedTeam almost two years ago we could clearly see where the internet was heading. We've since seen clients with over 90% of their traffic made up of not just automation but highly intelligent bot swarms, capable of bypassing any traditional detection mechanism. The only way to combat a threat like this is to welcome it and learn from it. By constantly ingesting and evolving attack vectors from a hive mind of miners, a Bittensor powered defence has now become the only reliable source of protection on the internet.
Here is @const_reborn's full keynote at @ExploitSummit It is worth the 36 minutes, but I'll give you the cliff notes. > Affine 1, a model trained by anonymous miners entirely on Bittensor, matches the intelligence score of Anthropic's Opus 4.1. > 23 of 128 subnets now have paying customers > Gamma tokens are coming to turn subnet usage credits into something agents can buy from any chain. > "We're not afraid of hyper-intelligent bot swarms on the internet. We use them to optimize our technology."
1
19
3
108
7,322
A DOJ asset seizure warrant unsealed this month is worth reading closely if you work in fraud and financial crime prevention. The instructions to donors were specific: don't use Binance, use these five apps instead, enter fictitious data as the recipient, expect the wallet address to change after every transfer. Every one of those instructions targets the same layer: Identity. Change the name, change the wallet, change the app, change the jurisdiction... What's harder to make disposable is the machine executing transactions. Threat actors are accustomed to operating a small set of devices across countless accounts and false identities. The response to this case will mostly be more KYC and tighter rules on issuers. Both are reasonable. However the people writing those instructions already told us which control they were designing around, and it wasn't the device. Greater compliance requirements need to be put in place that incorporate device-level controls. @innerworksID @_redteam_ @oscar_hayek
1
2
5
303
Over a billion dollars has already been invested into the “Know Your Agent” category. There are exceptions who are building critical updated auth infrastructure - but the absolute majority of these new players lack substance and are avoiding tough R&D. The agent economy requires new security primitives. Incumbent solutions to establish very baseline truths about a user eg what device are they using? Where are they truly located?…These are not resistant to agentic threats and aren’t cost efficient to deal with scale of use cases like agentic payments. #agenticpayments #agenteconomy #security #authentication
3
178
Imagine a bank with a fancy new security system… Then someone finds a spare key to the back door lying in the car park. They walk in. None of the clever equipment ever gets tested, because it was all pointed at the front door. That’s what actually happened. The agents didn’t invent a brilliant new hack to get inside Hugging Face. One of them found working login details already sitting out in the open on the internet and posted them to the group. Once they were in, the activity was exceptionally sophisticated, escalating access, setting up a fleet that respawned when it was deleted. The point of access could have been secured. What this case shows is how much depends on being able to trace agent activity back to the authorising device and user. dwarkesh.com/p/openai-huggin… #DeviceIntelligence #AgenticSecurity @oscar_hayek @_redteam_ @innerworksID
3
153
We are living through a strange moment in the history of financial security online. On one side sits a new industry, well capitalised and genuinely sophisticated, built to answer a question that would have sounded absurd not long ago: is this face real. Liveness. Injection detection. Deepfake forensics. We have put a generation of talent to work on drawing the distinction between human and synthetic. On the other side, the thing actually standing between an attacker and your cash tonight is a six digit code sent over a signalling protocol standardised in the 1980s, to a phone number a stranger can take over with a convincing story and ten minutes on the phone to a call centre. Financial Services platforms are running the most advanced fraud detection ever built on top of the weakest primitives ever. Both things are true at the same time, in the same login flow, at the same institutions. A password can be bought for a few dollars. A code can be intercepted. A face can be synthesised. Device signals can be spoofed too, and a spoofed device will pass a shallow check without much trouble. The critical difference is that a device is not one signal, it is a stack. What the browser reports about itself, how the graphics hardware actually renders, how the network stack negotiates a connection, how long packets really take to arrive. An attacker controls the top of that stack easily and the bottom of it barely at all. #DeviceIntelligence is the starting point for security. @_redteam_ is the R&D engine driving the future of it. @oscar_hayek @innerworksID
1
1
7
258
Oliver Quie retweeted
The Immune System for the Internet. Releasing next month.
Defence in this era must be AI driven, the absolute baseline is adapting defences faster than attacks are being produced. Anything slower than this leaves you in a constant state of degradation. What we’ve built on @_redteam_ is an immune system for this problem. The attacks we ingest are evolved inside the system into complete novel vectors that cannot be produced anywhere else. Every one of these gets patched before an external attacker has the means/ability to create it, let alone release it. bittensor:native #SN61
1
11
3
86
5,897
Couldn’t be more excited for @oscar_hayek to showcase the technology emerging from the incredible @_redteam_ community. Decentralized intelligence is winning.
Incredibly excited to join the @YumaGroup subnet spotlight to chat about @_redteam_ this Wednesday
5
396
Amazing work at @_redteam_ Excited for what’s ahead
🔍 ab_sniffer_v3 - Live Today The AB Sniffer challenge just leveled up. Building on v1 and v2, this new iteration introduces detection for an additional automation framework: Camoufox. This brings the total to 7 frameworks in scope. To raise the difficulty further, the repeated framework count has been increased from 3 to 4, meaning you’ll need to prove consistency across more runs. Check out the PR here and start mining: github.com/RedTeamSubnet/Red…
2
4
5
577
Out of thousands of startups globally, @innerworksID has placed first in the @SantanderX Global Award. A belated but timely announcement, now that Santander X is accepting applications for 2025! This is a watershed moment for AI Safety getting serious recognition from mainstream banking. @innerworksID is defining an entirely new frontier: Synthetic Threat Intelligence. While the industry debates AI safety in abstract terms, we're addressing the concrete reality - adversarial AI systems that generate perfect simulations of authentic human users to commit financial fraud online. Our R&D engine, @_redteam_ is absolutely mission critical to identifying these simulations - decentralised training is a totally new paradigm to build detection systems that thrive in the face of rapidly evolving threats. When a top tier bank validates this threat landscape, it signals a fundamental shift that protecting against Gen AI threats is transitioning from a specialised capability to critical infrastructure. We have had an amazing journey with @Santander, from meeting @J_Ennis & @GNev2 at the UK Finals - to travelling to Madrid for the Global Award and meeting leaders from @Google, @Telefonica and @IBM Thank you to everyone who supported us and good luck to the firms applying this year. youtu.be/cJGZOqM1FUg #AISafety #SyntheticThreatIntelligence #Cybersecurity #FraudDetection #TransactionSecurity @oscar_hayek
1
3
3
14
4,647
The dirty secret of Web3? Most protocols are flying blind - they have no idea how many authentic humans are using their platform vs bots, scammers, and bad actors. Had a fascinating discussion with Jason Jiang of @CertiK, Denys Ivanov of @hackenclub , and Thomas Menard at @proofoftalk today about why proactive security audits are an existential for the industry. Shoutout to @XVentures_Fund for bringing together the right voices on this critical issue. #Web3Identity #SecurityFirst #ProofOfTalk #DecentralizedTrust @_redteam_ @oscar_hayek @innerworksID
1
3
9
540
This is decentralised threat intelligence in action. A global hive of engineers outpaces any centralised R&D team. If a GenAI threat comes for your business, you’ll want the swarm on your side. @oscar_hayek @innerworksID @_redteam_
🚨 Two new adversarial challenges are landing on SN61 RedTeam — designed to push both your bots and your detection systems. First up: 🎯 humanize_behavior_v4 (June 9) A more advanced detection system is now in place — including reverse engineering of prior challenge (v3) responses and additional behavioural obstacles. Next: 🔍 ab_sniffer_v1 (June 10) Detect what’s meant to stay hidden. Build a Node.js SDK that can reliably sniff out which automation framework is driving the session — selenium, seleniumbase, patchright, or nodriver. You’ll need 100% accuracy on at least 2 out of 4 to qualify. Both challenges are live now as PRs — review and start prepping your submissions: 👉 github.com/RedTeamSubnet/Red… 👉 github.com/RedTeamSubnet/Red… RedTeam is a Bittensor subnet where real-time, adaptive defence gets rewarded. Let’s see if you can keep up. #Cybersecurity #Bittensor #RedTeam #BotDetection #AI #Automation
1
11
844
Oliver Quie retweeted
Bots now make up 60–70% of all internet traffic-and the number keeps rising. As automation outpaces human activity, static defenses just can’t keep up. @_redteam_ , a decentralized, always-on cybersecurity protocol. Any company can deploy real-world challenges on our subnet tapping into a global network of ethical hackers to solve their toughest security issues. This is the future of adaptive, accessible defense. #Cybersecurity #Bittensor #RedTeam #AI #BotDetection
1
8
1
23
1,738
Please get me a filter by humans button @elonmusk. The world would be a better place.
2
379
Oliver Quie retweeted
The narrative of DeepSeek is probably the biggest validation of deAI essentially to date. There is no pretraining moat, there’s probably little to no data moat longer term. AI is a feature that cannot be maintained by Big Tech as prop products. AI, like all tech and content, wants to be open and free. Trying to capture and lock in all the users with shiny UX, and OS integrations, failed. The next viable business model for AI is then monetization, governance, and public ownership of its production process. To my thesis, that’s exactly what the deAI stack provides, starting with decentralized training.
18
18
4
100
12,316
We can’t wait to share even bigger news from @_redteam_ very soon. @gTAOventures @oscar_hayek @innerworksID
🚀 Exciting News from Subnet 61‼️ We're thrilled to announce the release of our latest blog post: "Unlocking Value: Monetising RedTeam's Outputs and Challenges". This in-depth article explores three key monetisation opportunities within our ecosystem: - Miner Outputs - Challenge Detection Components - Third-Party Commissioned Challenges medium.com/@TheRedTeam/unloc… Stay tuned for our next blog post, where we'll dive into how our Alpha token and dynamic TAO update will further enhance these opportunities.
1
3
271
📢New Bot Detection Challenge on sn61 #Bittensor @_redteam_ Miners are tasked to bypass traditional behavioural analysis, where the mouse movements of the bots are tracked and analysed for non-human / anomalous behaviours. Read more in the blog here👇 medium.com/@TheRedTeam/bot-c…
3
5
449
The internet is no longer dominated by humans—most online traffic is already bots. What was once a warning about fraud has transformed into the rise of something extraordinary: the powerful emergence of the agent economy. The full potential of that transformation is at risk, and I want to share more about how we are helping. Please meet @innerworksID and @_redteam_ . 🚀
1
1
5
629
The takeaway: The companies that prepare for a majority-agent internet will lead the next wave of innovation. To thrive, you need: 1️⃣ Sophisticated fraud detection to protect your systems 2️⃣ Seamless infrastructure to enable trust 3️⃣ Proactive defenses to stay ahead of threats
1
1
100
in my next thread I’ll share more about how decentralized AI is the backbone of everything we are building. #Cybersecurity #AI #FutureInternet #ProofOfHumanity #AgentEconomy @gTAOventures @YumaGroup
2
7
324