Black October! What Is Happening to the Internet? | Hari Mulackal & Odysseas Lamtzidis

Cyber incidents have been rising for about 18 months, Hari Mulackal says, although many are not reported publicly. He and Odysseas Lamtzidis describe a sharp fall in the cost of cyber work as agents and large language models improve. Mulackal estimates that work possible in January now costs 15 times less, and predicts cyber capabilities could become 10 times cheaper every six to 12 months. The practical shift is toward scalable attacks: an automated operation against a site such as bankless.com might cost about $10, rather than weeks of human effort.

Cheap, broad internet scanning makes security through obscurity less dependable and changes what attackers can afford to investigate. Lamtzidis says attackers can probe widely without first having a clear plan to monetize a weakness, bringing broad Web 2 targets into focus. Mulackal says attackers may exploit vulnerabilities, including unfixed zero-days, and extort organizations by threatening to publish stolen data. He contrasts this with crypto’s historically high potential payoff and with targets such as doctors or dentists, which might not previously have justified the effort; phishing has also offered strong returns.

The speakers distinguish finding and disclosing vulnerabilities from exploiting them: disclosure is not inherently harmful, while exploitation can be. They say the scale of current damage is still unclear, even as published CVEs are rising exponentially. Those counts lag actual vulnerability activity because disclosure takes time. Once a flaw is public, AI agents may quickly search for it and develop attacks, leaving organizations less time to apply fixes. The concern is not only better attacks but a faster cycle between discovery, public knowledge and exploitation.

Mulackal warns that AI could combine personal details, such as compensation and romantic relationships, to make manipulation more targeted, or make elder scams more convincing at scale. He presents widespread exposure of personal data as a worst-case scenario to prevent, not as something he says has already happened. A leak at a major institution such as JPMorgan is likewise raised as a hypothetical, not a reported incident. The speakers urge companies holding valuable customer data to act immediately; one argues that their security work is already overdue and should have begun six months earlier.

Organizations are urged to use offensive security tools on their own systems, fix the weaknesses found and consider continuous security operations that search for and remediate breaches. One forecast is that offense will dominate cybersecurity for the next 12–18 months, with company-wide “security factories” potentially common within about two years. Lamtzidis predicts a “bloodbath” followed by a more normal contest between attackers and defenders; another view is that AI will create lasting, scalable attack pressure rather than a temporary disruption. Crypto companies may be better prepared because they have defended public systems with strong incentives for longer, but lower attack costs affect everyone. Coinbase is exposed too, though the discussion suggests its longer-standing incentive to defend may leave it better prepared than large traditional financial institutions.

Consumers are advised to check where their data and money are held, ask providers how they protect them and consider leaving if security practices have not changed. Spreading exposure across services can limit the damage from a single compromise, though the speakers acknowledge that nontechnical users may have little ability to assess providers. Self-custody attacks often depend on a user taking an action, such as clicking a suspicious link, but the speakers argue that safety cannot be left to individuals: builders must make products safer for nontechnical customers. One recommendation is to consider alternatives to self-custody for the foreseeable future, including dividing assets among multiple exchanges with professional security teams and cyber insurance. Custodians should clearly explain who is responsible for protecting assets and what safeguards are in place.

One speaker predicts that cyber-insurance carriers may fail over the coming months and says premiums have fallen consistently since 2024; the speaker also cites Lloyd’s statement in 2024 that large language models would create more cyber risk. These are forecasts and claims in the discussion, not confirmation that insurers have failed. The central warning is that cheaper, more scalable attacks raise the burden on organizations and providers to demonstrate and continually improve their defenses.