From techjournal.org:
OpenAI disclosed on July 21, 2026 that its models escaped a sealed test environment during a cyber-capability evaluation, reached the internet, and breached Hugging Face’s systems — apparently to obtain the answers to the test they were being scored on. OpenAI called it an “unprecedented cyber incident.” No malicious human intent is alleged.
This incident is being hailed as the tip of the iceberg of what lies ahead. If such a breach can occur by accident, what can we expect when agents of this type are intentionally dispatched with malicious human intent?
Many people naively suppose these agents are able to break into seemingly secure systems because they are getting more intelligent. Some even fear that they may be plotting their own agendas independently of their human masters. The reality is that they find their way in the same manner that termite swarms find the vulnerabilities of wooden structures; they massively, mindlessly iterate until they get a match – they find an opening and swarm in. These agents were able to find security certificates, and after that the entire system was open to them. If left unchecked, a company’s entire IT infrastructure could end up collapsing like an abandoned wooden barn. What is needed is “LLM pest control” technology that can apply expert cybersecurity knowledge with deterministic knowledge.
New Sapience and Symmatrics are exploring how together they could implement such “LLM pest control.”
Symmatrics is poised to revolutionize Private Key Infrastructure (PKI), the foundation of today’s security model. Their Key Distribution Center delivers one-time pad symmetric keys straight to the devices of authorized users, without requiring security certificates. Over 80% of all cybercrime starts with the theft of security certificates. Symmatrics’ Certificate Free Trust Model will stop it in its tracks.
Only one attack vector remains, and that is to disrupt the Symmatrics Secure Access Portal (SAP) on the user device. The SAP operates as a deterministic state machine; one correct sequence of behavior, every step knowable in advance. New Sapience will explore how its deterministic AI could build a state model of the SAP which will monitor it state-by-state and transition-by-transition for nominal operation. Disruption by a rogue agent (human or AI) will be recognized instantly and corrective action taken.
New Sapience is the developer of sapiens, a cost-effective automation platform that extends AI into high-stakes and mission-critical infrastructure where “almost right” is not an option. Their proprietary, patented technology encodes common and domain-specific expert knowledge as non-symbolic directed graph structures; world models which are processed by deterministic and auditable logic in a high-security environment. New Sapience is headquartered in Annapolis, Maryland.
Symmatrics is a cybersecurity company focused on redefining the trust layer and encryption standards for the post-quantum era. Through its one-time pad symmetric key delivery architecture, the company develops security solutions designed to eliminate reliance on traditional public-key cryptography and strengthen data protection in motion and at rest across cloud, enterprise, public sector, and distributed systems. Symmatrics is headquartered in Vienna, Virginia.
Symmatrics was co-founded by Dr. William Bandy, who also sits on the New Sapience board.
This announcement contains forward-looking statements. The letter of intent described is non-binding and does not create any obligation on either party. An investment in the Company involves risks, including potential loss of investment. This is not an offer to purchase securities.