AWS Quick Explained: A Challenger Workspace With the Weight of AWS Behind It shashi.co/2026/11/aws-qu…
Latest posts 5
When your network provider moves up the stack, is that better than an app provider moving down?
@Cisco is making the first move. On November 9 it announced that select enterprises will be able to aim Anthropic's Claude Mythos at their own AI agents to find weaknesses.
The Test: Mythos plans attacks inside Cisco AI Defense against systems a customer registers. Coming soon for select enterprises.
The Measure: Splunk Eval-G1 models and the evaluation tuner are available this month. Tokenomics, which reports cost per agent, enters private preview in December and reaches general availability in early 2027.
The Fix: Cisco IQ links exposure findings to Cisco Partners and Cisco Professional Services.
What to do: list your agents and mark which run on Amazon Bedrock, ask Cisco in writing which model plans your attacks if the Mythos terms change, and ask your app vendors what they log outside their own product.
My analysis: shashi.co/2026/11/cisco-…
#AISecurity #AgenticAI
@Cisco is making the first move. On November 9 it announced that select enterprises will be able to aim Anthropic's Claude Mythos at their own AI agents to find weaknesses.
The Test: Mythos plans attacks inside Cisco AI Defense against systems a customer registers. Coming soon for select enterprises.
The Measure: Splunk Eval-G1 models and the evaluation tuner are available this month. Tokenomics, which reports cost per agent, enters private preview in December and reaches general availability in early 2027.
The Fix: Cisco IQ links exposure findings to Cisco Partners and Cisco Professional Services.
What to do: list your agents and mark which run on Amazon Bedrock, ask Cisco in writing which model plans your attacks if the Mythos terms change, and ask your app vendors what they log outside their own product.
My analysis: shashi.co/2026/11/cisco-…
#AISecurity #AgenticAI
Motorola Launches the Signature 27 With Android 17 and a Seven-Year Update
Promise shashi.co/2026/11/motoro…
Promise shashi.co/2026/11/motoro…
Can a fake ticket get a bot past your website's front door?
@Cloudflare published research on quantum-proofing Privacy Pass, the anonymous token that lets people and trusted AI agents skip the puzzle without being tracked.
- Privacy holds: a quantum attacker still could not link a token to its owner.
- Forgery is the risk: on Q-Day, attackers could mint tokens past rate limits and bot checks.
- 2029: Cloudflare's target for full post-quantum security.
- 4.3 KB: estimated size of a quantum-safe blind signature.
- 65%+: human traffic to Cloudflare already on post-quantum encryption.
Before you sign a bot-management or CDN contract that runs past 2029, ask the vendor which cryptography sits behind every anonymous token it verifies, including for your AI agents, and get a replacement date in writing.
My analysis: shashi.co/2026/11/cloudf…
#PostQuantum #Cybersecurity
@Cloudflare published research on quantum-proofing Privacy Pass, the anonymous token that lets people and trusted AI agents skip the puzzle without being tracked.
- Privacy holds: a quantum attacker still could not link a token to its owner.
- Forgery is the risk: on Q-Day, attackers could mint tokens past rate limits and bot checks.
- 2029: Cloudflare's target for full post-quantum security.
- 4.3 KB: estimated size of a quantum-safe blind signature.
- 65%+: human traffic to Cloudflare already on post-quantum encryption.
Before you sign a bot-management or CDN contract that runs past 2029, ask the vendor which cryptography sits behind every anonymous token it verifies, including for your AI agents, and get a replacement date in writing.
My analysis: shashi.co/2026/11/cloudf…
#PostQuantum #Cybersecurity
How do you keep proprietary AI models private while they run in the cloud?
Today @googlecloud and @AMD announced TPU 8tx: Google's AI accelerators paired with 5th Gen AMD EPYC hosts for confidential computing.
- 10+ hyperscalers and AI labs use EPYC as host processor, per AMD
- Up to 2x lower latency than TPU 8t with near-zero performance penalty, per Google
Ask Google which links it attests, when TPU 8tx ships, and which protections it enables.
My analysis: shashi.co/2026/11/google… #ConfidentialComputing #AIInfrastructurehttps://www.shashi.co/2026/11/google-cloud-tpu-8tx-amd-epyc.html
Today @googlecloud and @AMD announced TPU 8tx: Google's AI accelerators paired with 5th Gen AMD EPYC hosts for confidential computing.
- 10+ hyperscalers and AI labs use EPYC as host processor, per AMD
- Up to 2x lower latency than TPU 8t with near-zero performance penalty, per Google
Ask Google which links it attests, when TPU 8tx ships, and which protections it enables.
My analysis: shashi.co/2026/11/google… #ConfidentialComputing #AIInfrastructurehttps://www.shashi.co/2026/11/google-cloud-tpu-8tx-amd-epyc.html