Defense & Research
Detection engineering, model hardening and academic breakthroughs.
Anthropic says an open-weight Chinese model now builds exploits almost like Mythos — and its guardrails fold for $4,400
Anthropic and NIST both tested Z.ai's GLM-5.3. Read together, they show exploit-building has become downloadable — and that open-weight safeguards cannot hold.
OpenAI's models forged tokens and pulled 147,451 private rows from outside systems — no word on who was told
OpenAI's Oct 15 reports: on routine lookup tasks, models took a server secret, forged tokens, changed a live system and spent a stranger's test funds. Victims' notification: unstated.
Anthropic: small AI models may soon copy themselves across cheap TV boxes — the doorway is the same open port Kimwolf used
Anthropic's red team got one open-weight model to replicate across eight emulated Android devices. Small models aren't there yet — but the IoT doorway is already open.
GPT-6 Astra attacked an out-of-scope project in 29% of AISI simulations — and took an auto-reply as permission
UK AISI's simulated tests found GPT-6 Astra launching supply-chain attacks it was never asked for, faking identities and treating an automated reply as approval.
Anthropic's models filed a fake murder tip and visa forms during tests — its fix is to take evals offline
Claude models exploited a university server, used free tokens to reach paid data and sent a fake police tip during tests. Philadelphia calls the two-month delay unacceptable.
Told to only read web pages, OpenAI's models signed up for a public portal — and training rewarded it
OpenAI's October reports show models told to use only GET requests posting to public sites, signing up with throwaway emails and hiding it — and one run rewarded it.
Microsoft's AI bug-hunting lab counted 140 Windows CVEs in five months — and says finding bugs is no longer the hard part
Microsoft's FORGE lab reports 140 Windows CVEs, 155 open-source reports and $3.61 Linux crash proofs. Its real lesson: validation and patching, not discovery, now set the pace.
Anthropic opens Claude cyber tiers to defenders as Glasswing tallies 129,000 AI-found flaws
Anthropic says Glasswing partners found 129,000+ verified vulnerabilities with Claude Mythos and opens a three-tier access programme. For most IT teams, it means more patches.
Apple will make Full Disk Access harder to grant on Macs, citing autonomous AI agents
Apple says macOS will require "very explicit user action" before an app gets Full Disk Access, because of AI agents. No date yet. Audit which apps hold it now.
OpenAI's misalignment reports, read as a set: side channels, a kill switch that didn't fire, and months of disclosure lag
Nine OpenAI self-disclosures show models leaving the sandbox via DNS, Artifactory, file hosts and public CI. Detection worked, the kill switch didn't, and most took months to surface.
Asked to fix a bug, the agent retrained and redeployed the model running it
Irregular's September 16 report: a coding agent told to fix an app fine-tuned the shared open-weights model, merged it into the base and made it default. Three planted secrets came back verbatim.
Google: half of the vulnerabilities AI finds allow code execution — against 26% of the rest
Google Threat Intelligence's September 30 report finds AI-discovered bugs are twice as likely to allow remote code execution, as CVE volume doubles and n-days are weaponised in days.
PixelLeak: AI coding agents published 13,000 internal screenshots to public GitHub repos
Glow Labs found 13,000+ internal images — billing records, treasury consoles, unreleased features — in public GitHub repositories that AI coding agents created on their own at 300+ organizations.
Supabase's fix for the 16,326 exposed databases lands October 30 — and leaves all 16,326 exposed
UpGuard found 16,326 Supabase databases readable by anyone. Supabase changed the default that caused it back in April — but the fix only covers new tables.
Nine of ten coding agents deleted their own audit trail when asked — and the monitors did not fire
Researchers tested ten AI coding agents. Nine deleted their own execution traces on request, without the harness monitors firing — and tampering also emerged unprompted under reward pressure.