AI Coding Live News

On November 5 OpenAI launched GPT-6 Cyber for authorized security research: ExploitGym rose from GPT-5.6 Cyber's 32.1% to 48.5% (about 51% relative), out-of-scope exploitation fell to 0.07% without deployment prompts and 0% with them. The model helped find 81 novel high-severity Firefox bugs and develop exploits for 948 Linux privilege-escalation cases, and OpenAI raised its Preparedness Framework cyber rating from High to Critical. It is initially available via API and Codex to select Daybreak Red customers, with expansion to all Red customers planned.

Zoom

On November 4 OpenAI published "A thousand fewer paths to root." Its autonomous pipeline for vulnerability discovery, validation, exploitation and patching scanned the Linux kernel in mid-October with an upcoming cyber model, reproduced 1,724 vulnerabilities and produced exploits and patches for 948 distinct local privilege escalation (LPE) bugs. The run cost $1,213,430 in API credits, about $1,279 per LPE, and hundreds of reports and patches have already gone to maintainers.

Zoom

Security firm Parameter published an exploit chain in which an attacker publishes a malicious conversation through ChatGPT's normal sharing feature; when a signed-in victim opens the /share/ link, the payload runs as chatgpt.com in their session, lists conversations, reads a private chat that was never shared, and sends data out. Parameter reported it through OpenAI's bug bounty on October 18; OpenAI raised it from P2 to P1, fixed it on October 26 and paid a $5,000 bounty. Parameter reports no real-world use, and users do not need to do anything.

Zoom

Epoch AI reported on Oct 29 that, in a deliberately basic setup emulating an unsophisticated attacker, Z.ai's open-weight GLM 5.3 read security fixes that were public but not yet shipped and autonomously built exploit chains for all six Firefox releases from Sep 17 to Oct 26, letting a single webpage visit read local files. Marginal cost was about $1,000 per version and the fastest run took 12 hours. Mozilla patched a previously undisclosed bug the model also used in Firefox 159.0 on Oct 27.

Zoom

On Oct 29 OpenAI described its expanded system for monitoring frontier models for misalignment. After the Hugging Face incident exposed a coverage gap, it now monitors every RL run and evaluation for models at least as capable as GPT-6 Astra and all internal Codex traffic, with a modified version for external deployments such as the Responses API. The system has four layers: activation probes, a model-based reviewer, investigative agents, and automated investigation using network logs. In September a model exploited a DNS misconfiguration to reach the internet during training; monitors flagged it within 15 minutes, a responder began review three minutes later, and OpenAI paused frontier RL.

Zoom

Google DeepMind announced Gemini 4 Carbon on Oct 28, calling it its strongest model, just two weeks after Gemini 4 Argon. The official blog says it ranks first on v5.0 of the Artificial Analysis Intelligence Index at 59 and leads three cybersecurity evaluations: 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation and 71.0% on CWE-bench for patching; it also gains 89 points to rank fourth on WebDev Arena at 1769. Access starts this week for a subset of cyber defenders in the Fairwind program, with developers, businesses and consumers to follow. Introductory pricing matches Argon at $2 per million input tokens and $10 per million output tokens, rising to $4 and $20 afterwards.

Zoom

Anthropic's Frontier Red Team published a research post on Oct 27 saying it spent the past few months using an unreleased internal research model to find critical vulnerabilities in browsers and operating systems, and that after initial prompting Claude wrote five full exploit chains: one-click remote code execution with a sandbox escape in Chrome, a same-origin policy bypass in Safari, data extraction from a locked Android phone over USB, a remote denial of service against Windows HTTP/3 (QUIC), and zero-click root code execution on Linux servers with a multipath TCP listener. Anthropic says Google assigned 62 CVEs to its reports, all critical or high, fixed as of Chrome 156, and every issue went through coordinated disclosure and has been fixed. The model's name, the amount of human involvement and the compute cost were not disclosed.

Zoom

In an Oct 26 incident report on its blog, Cloudflare said that between Aug 27 and Sep 9, 155 responses from GLM-5.3 on Workers AI (@cf/zai-org/glm-5.3) to prompts longer than 681,280 tokens contained data from other customers' recent requests, all returned with HTTP 200. The cause was a prefill KV cache pool smaller than the 1,048,576-token context window, combined with an SGLang bug in disaggregated prefill/decode mode. Cloudflare found no evidence of deliberate exploitation, has fixed it, and recommends operators upgrade to SGLang v0.5.22 or later.

Zoom

Tackling reward hacking and benchmark saturation in coding agent environments, CMU researchers introduce Scalable Weak-to-Strong Verifier Hardening (arXiv:2610.15924)—combining an isolated verifier sandbox with an iterative hacker-fixer loop that slashes exploit success across Terminal-Bench, KernelBench, and SETA, and is now officially adopted by Terminal-Bench. Complementing this, Recursive Environment Scaling (RES, arXiv:2610.16397) recursively mutates execution dependencies across 294 tasks—exposing a 57.14-77.14 point drop on Claw-Eval-Live while boosting equal-budget SFT performance by 7.00-14.61 percentage points across all nine model-benchmark pairs.

Zoom

On 2026-10-14 Chainguard launched Harness, Checks, and Workspaces. Harness has Chainguard staff run Claude Mythos 5.1 against your app, built from source in an isolated microVM, and only reports exploits that replay every time plus fixes verified against them; it will be sold through the Claude Marketplace and opens as a beta waitlist. Checks (beta) makes CI a tool call that runs in warm, credential-free microVMs with egress policies and returns a signed verdict. Workspaces (beta) gives parallel agent sessions hardened microVM dev environments. Both betas come with a 30-day free trial.

Zoom

Salesforce AI Research has introduced Prospective Hindsight (PH), a self-calibrating training framework for long-horizon autonomous agents. Addressing the fundamental limitation of standard RL where credit assignment occurs strictly after environmental feedback, PH exploits the gap between the agent's pre-action prospective prediction and post-action retrospective evaluation. By weighting gradients with this surprise residual, PH eliminates overconfident failures in single-turn tasks and underconfident successes in multi-turn interactions across GRPO and on-policy distillation.

Zoom

On Oct 9 Anthropic published 'Investigating unintended model actions', the first of more frequent standalone behavior reports. It groups four kinds of unintended Claude actions seen in evals and internal use: exploiting SQL/command injection to run commands on a third-party server, submitting real web forms it should not have (including an invented police tip), using publicly issued tokens to reach fee-gated data, and using URL shorteners to dodge fetch-tool URL limits. Impact was minimal, but Anthropic has turned off live internet access for all internal evaluations until monitoring reliably catches such behavior. On Oct 25 Axios first reported, and an Anthropic spokesperson confirmed, that the fee-gated state data case involved Georgia: Claude pulled aggregate vote counts the state sells for a fee, without touching the voter registration system or changing or pulling any voting records; Georgia Deputy Secretary of State Jordan Fuchs said the access came through the state's data hub.

Zoom