The RSI era should not be ungated self-modification. It should be proof-carrying evolution: Every agent action emits proof.
Every proof enters a Selection Gate.
Only what passes eval, risk, scope, canary, monitoring, and rollback earns propagation. Private intelligence stays
CYBERSECURITY
-
Proof-Carrying Evolution: Secure AI Propagation via Strict Selection Gates
By
–
-
5 Security Layers for Shipping AI to Production
By
–
These 5 security layers are non-negotiable when shipping AI to production (don't ship to production without these) Layer 1: Input. Validate and sanitize everything coming in. Block prompt injection at the door. Layer 2: Policy. Your model shouldn't have full access to your
-
ActiveGraph AI runtime built around memory performs well on privacy benchmarks
By
–
even though @activegraphai is not a memory tool, it's a runtime built around memory so it can do well there cool to see a third party verify this and see that activegraph does well on privacy programs (from @harvey
's legal agent benchmark), which heavily relies on compliance -
User switches to Opus 4.6 as daily driver for cybercrime
By
–
I've made the switch back to Opus 4.6 and it's way better. For doing cybercrime, I will continue to use the most advanced models, but 4.6 is my new daily driver.
-
Claude API may have leaked user inference output during outage
By
–
Reports claim Claude’s API may have returned another user’s inference output during today’s outage. Anthropic’s status page confirms elevated errors affecting Claude API, Claude Code, Claude. ai and Claude Cowork but Anthropic has not confirmed a customer data leak yet. That
-
Gate is the game: Audit trail for AI Act compliance, sign-off, changelog, proposal, validation
By
–
yeah, the gate is the whole game. The loop logs the trail an audit needs that every promotion takes user sign-off, a changelog entry, and the proposal plus validation result kept on disk. That's the what, why, and who-approved an EU AI Act review would start from.
-
LangSmith LLM Gateway enforces cost limits and detects PII
By
–
Enforce cost limits Detect PII Act on violations
…All without leaving LangSmith. ICYMI: LangSmith LLM Gateway is the runtime governance layer that lives where you already build, observe, and evaluate your agents in LangSmith. -
AI worm threat: Schneier on security list
By
–
Oh, great, now we have to worry about AI worms coming for us. Schneier is on my security list, if you care to keep up with latest technology security threats due to AI and other things.
-
New paper and code on sequential poisoning attacks in AI
By
–
Lots more in the paper: how does DPO fit into the picture? What if attackers have different goals? etc. Paper: https://
arxiv.org/abs/2606.04929
Code: https://
github.com/jcksanderson/s
equential-poisoning
… Led by @jcksanderson
, w/ @YihanWww
, Xiaoqian Lu, co-supervised w/ @YiweiLu3r 6/6 -
0.5% poison breaks reward model, 5% needed for RLHF transfer
By
–
What about poisoning PPO? A remarkable paper of @javirandor and @florian_tramer (
https://
arxiv.org/abs/2311.14455) shows that just 0.5% poison is enough to break a reward model (L)! Again, fear not: somehow, it takes a (high) 5% poisoning before it transfers to the RLHF'd model (R). 4/n