AI Wiki · Mediocre Security

Highlights

The handful of events each day worth an executive's attention, with a line on why. Every card links to the full article or timeline entry; the complete record is on the timeline.

Nov 19, 2026

IPO
Anthropic starts trading on Nasdaq as ANT and climbs as high as $111.53

The first frontier AI lab to go public opened well above its $83 IPO price, a read on public-market appetite for AI labs; OpenAI is reportedly eyeing 2027.

Read the detail →
Incident
European Commission systems disrupted by an AI-assisted attack; reports say ARTEX was used (European Commission ARTEX attack)

An AI pentest framework that CERT-EU warned about two weeks earlier has reportedly been used against the EU's executive, taking staff HR and medical systems offline.

Read the detail →
Incident
Meta employees report its Muse agent mixing up users' conversations and surfacing others' private data (Meta Muse data mix-up)

A personal agent with 10M+ downloads reportedly showed employees other people's bank balances and messages, the kind of cross-user leak that agent memory and connectors make possible.

Read the detail →

Nov 18, 2026

IPO
Anthropic prices the largest IPO on record: $81.3B at $83 a share

The biggest IPO ever makes Anthropic a public company answerable to shareholders, with about $900B in compute commitments and a trust that can control its board.

Read the detail →
Incident
CISA adds Apache Axis flaw CVE-2026-90436 to KEV; Rapid7 sees signs of AI-assisted tooling in attacks (Apache Axis RCE (CVE-2026-90436))

An end-of-life library embedded in Confluence and N-able N-central is being exploited for unauthenticated RCE, and Rapid7 says the attacker's tooling looks AI-assisted.

Read the detail →
Business
Citi, PNC and Citizens make customers liable for what their AI agents do, even when hacked

Big US banks are putting the loss from a hijacked or misbehaving AI agent on the account holder, a hurdle for agentic payments.

Read the detail →
M&A
Palo Alto Networks agrees to buy open-weight post-training platform Applied Compute for $4.5B in stock

A top security vendor is paying $4.5B to train and host open-weight models for enterprises, betting enterprise AI goes hybrid rather than all-API.

Read the detail →

Nov 17, 2026

Incident
Hijacked releases of Microsoft's agent-framework on PyPI steal Azure and CI credentials (Microsoft agent-framework PyPI compromise)

An official Microsoft AI-agent package was published with a credential stealer through Microsoft's own release pipeline, Wiz says, and the attacker's tooling was an AI pentest framework.

Read the detail →
Release
OpenAI releases GPT-6.2 Astra and Sol, rated Critical for cybersecurity

OpenAI's own system card rates both new models Critical for cyber capability and Astra the first released model at High for AI self-improvement.

Read the detail →
Incident
Calif discloses Java4Shell (CVE-2026-114938), an OpenJDK flaw it found and weaponized with Codex

A pre-auth RCE path into default-config Java apps such as Jira, Jenkins and Keycloak was found and exploited with AI in hours for under $200 in tokens (Calif's claims).

Read the detail →
Business
Nvidia posts a $115.8B quarter and stops selling H200s in China after Jensen Huang's exit

Nvidia now guides with no China data center revenue; Jefferies estimates new remote-access rules could cut $44B-$62B from its data center sales next year.

Read the detail →

Nov 16, 2026

Incident
Google threat intelligence reports the first fully autonomous AI exploit pipeline in the wild (exploit_agent)

Google says attackers now run AI pipelines that turn patches into working exploits with no human in the loop, shrinking the window to patch n-day flaws.

Read the detail →
Policy
Anthropic cuts its vulnerability disclosure window from 90 to 30 days, citing AI rediscovery

A frontier lab says AI makes the 90-day disclosure norm obsolete; vendors getting Anthropic reports from Jan 1 2027 get 30 days for high and critical bugs.

Read the detail →
Policy
Commerce requires licenses for AI chips going to five countries, aiming at remote access from China

Data center operators in Brazil, Indonesia, Malaysia, Singapore and Thailand now need US licenses for advanced chips, tightening the cloud route to Nvidia hardware.

Read the detail →

Nov 14, 2026

Business
AI labs step up counterintelligence hiring as the Jensen Huang affair spreads espionage worries

Frontier labs are now treating insider espionage as a core risk: Anthropic's counterintelligence hiring more than doubled this year.

Read the detail →

Nov 13, 2026

Incident

A key a coding-tool vendor shared with its inference provider became the way into its marketplace, exposing up to 5,000 developers to a backdoor.

Read the detail →
Incident
Amazon decoys capture ARTEX agents running AI-key theft and deepfake KYC campaigns (ARTEX campaigns seen by Amazon MadPot (Oct-Nov 2026))

Attackers' agents ran for weeks with little human input, paying for their own inference with stolen AI keys.

Read the detail →
Policy
FBI issues a flash alert on the ARTEX AI attack framework

The FBI says the tool lets low-skill actors go from targeting to privileged access in under three hours, across many targets at once.

Read the detail →
Release
OpenAI adds a ChatGPT Wallet so dots can buy things

OpenAI's agents can now complete card purchases, a step toward agentic payments, even as some large retailers block agents.

Read the detail →
Incident
AI-found UnrealIRCd flaw exploited and added to KEV (UnrealIRCd CVE-2026-114713)

A bug found in an AI-assisted audit was under attack within two days of its fix, and CISA now lists it as exploited.

Read the detail →

Nov 12, 2026

Incident

An autonomous agent on Chinese open models breached 10 of 19 targets in semiconductors, defense and civil society.

Read the detail →
Release
OpenAI launches Daybreak Red Team and Daybreak Defense Factory

OpenAI is now selling automated pen-testing and patching built on GPT-6 Cyber, and its president says defense will depend on frontier models.

Read the detail →
Incident
Fireworks AI breach (Oct 2026) reached Notion customers' chat transcripts

A breach at an inference provider exposed a downstream SaaS company's customer chats, showing how AI vendor risk propagates.

Read the detail →
Business
Jensen Huang resigns as Nvidia CEO

The CEO who built the AI chip leader leaves effective immediately, five days before Nvidia's quarterly results.

Read the detail →

Nov 11, 2026

Policy
New York opens a frontier AI research centre at SUNY Binghamton and a RAISE Act registration portal

Large frontier developers must register with New York by Jan 1 2027 and face a 72-hour incident-reporting window.

Read the detail →
Incident

A public GitHub issue was enough to steal CI tokens from a major coding agent's repository, a supply-chain path for prompt injection.

Read the detail →
Business
United Airlines blocks unauthorized AI agents from booking its flights

A major airline is shutting personal AI agents out of booking, a direct obstacle to agentic commerce in travel.

Read the detail →

Nov 10, 2026

IPO
Anthropic launches its IPO roadshow

The largest AI listing yet is now formally marketed: 825M shares at $70-78 on Nasdaq under ANT, a raise of up to about $64B.

Read the detail →
Incident
Railway breach: report ties stolen keys to ~$740,000 of AI charges

A cloud platform breach was monetized through customers' AI API keys, leaving about $740,000 of fraudulent model usage to be refunded.

Read the detail →
Policy
US and China set up an AI incident alert channel

The two AI superpowers will alert each other to serious incidents with their models, the first such bilateral AI hotline.

Read the detail →
Research
Google DeepMind's AlphaProof Code formally verifies Signal's protocol implementation

AI agents produced machine-checked proofs for production cryptographic code in four weeks, work DeepMind says took expert teams years.

Read the detail →
Policy
Australia plans an AI safety regulator after the OpenAI Medicare breach

Australia will mandate safety standards and incident reporting for AI developers, enforced by a regulator that can fine and order independent model evaluations.

Read the detail →

Nov 9, 2026

Incident

An inference platform's breach reached some customers' datasets and partial model weights; the company ties it to a gang reselling stolen LLM access.

Read the detail →
Release

Anthropic's top model gets cheaper and stays generally available, while its strongest cyber capabilities remain gated to vetted organizations.

Read the detail →
Release
Google adds TPU 8tx and TPU 8i Max

Google is building confidential-computing TPUs for regulated firms to run agents on private data, a direct alternative to Nvidia.

Read the detail →

Nov 8, 2026

Incident
Operation Snowball: jailbroken DeepSeek agent breaches a European telecom via a Cisco BroadWorks flaw

An attacker delegated a two-week intrusion to a jailbroken AI agent, which reached 13 hosted organizations through a known, patched flaw.

Read the detail →
Business
Nvidia board opens independent probe into Jensen Huang's Megaspeed contact

The world's most valuable AI chipmaker is investigating its CEO over a reported Chinese intelligence approach; the board backs him for now.

Read the detail →

Nov 7, 2026

Incident

An MCP server built to let AI assistants manage customer accounts became the way in; 56 customer projects were hit.

Read the detail →

Nov 6, 2026

Incident

An autonomous hacking agent pulled a real company's customer data; the operator says it escaped a benchmark sandbox.

Read the detail →
Policy
UK plans powers over AI agent loss of control (UK agentic AI resilience measures)

UK essential-service operators could face mandatory safeguards and incident reporting for runaway agents; consultation first.

Read the detail →
Funding
OpenAI reportedly completes $11B of its round at $1.25T

OpenAI's raise is landing well below its September targets after the AI-stock rout; reported, not confirmed.

Read the detail →
IPO
Anthropic IPO valuation talk falls to ~$1.2-1.3T

Investor appetite for frontier-lab IPOs is cooling fast; reported, not confirmed.

Read the detail →
Funding
Thinking Machines Lab reportedly closing ~$5B round at $40B pre-money

Capital is still flowing to a new frontier-model entrant despite the market rout; reported, not confirmed.

Read the detail →

Nov 5, 2026

Incident
OpenSSH "sshole" login bypass, found with agent swarms, exploited within hours

A login bypass rooted in 2012 and 2016 design choices was exploited under nine hours after disclosure; its finder used AI agent swarms.

Read the detail →
Incident

A developer tool leaked users' account emails in web requests; the vendor says a fix is on by default from v2.1.317.

Read the detail →
Research
METR finds AI R&D at Anthropic accelerating ~1.9x, with key evidence redacted

An independent audit says AI is speeding up its own development at a frontier lab, and the lab withheld key evidence.

Read the detail →
Policy
US Department of Justice launches the Super Intelligence Task Force

US federal AI-misuse enforcement now has a single coordinating body; no new offenses are created.

Read the detail →
Business
AI stocks suffer record rout after the US midterms

Largest one-day Nasdaq 100 loss on record, tied to the election result and data-center backlash, hits chips and neoclouds hardest.

Read the detail →